Tag Archives: Publications

The Question of Quantum Supremacy



Quantum computing integrates the two largest technological revolutions of the last half century, information technology and quantum mechanics. If we compute using the rules of quantum mechanics, instead of binary logic, some intractable computational tasks become feasible. An important goal in the pursuit of a universal quantum computer is the determination of the smallest computational task that is prohibitively hard for today’s classical computers. This crossover point is known as the “quantum supremacy” frontier, and is a critical step on the path to more powerful and useful computations.

In “Characterizing quantum supremacy in near-term devices” published in Nature Physics (arXiv here), we present the theoretical foundation for a practical demonstration of quantum supremacy in near-term devices. It proposes the task of sampling bit-strings from the output of random quantum circuits, which can be thought of as the “hello world” program for quantum computers. The upshot of the argument is that the output of random chaotic systems (think butterfly effect) become very quickly harder to predict the longer they run. If one makes a random, chaotic qubit system and examines how long a classical system would take to emulate it, one gets a good measure of when a quantum computer could outperform a classical one. Arguably, this is the strongest theoretical proposal to prove an exponential separation between the computational power of classical and quantum computers.

Determining where exactly the quantum supremacy frontier lies for sampling random quantum circuits has rapidly become an exciting area of research. On one hand, improvements in classical algorithms to simulate quantum circuits aim to increase the size of the quantum circuits required to establish quantum supremacy. This forces an experimental quantum device with a sufficiently large number of qubits and low enough error rates to implement circuits of sufficient depth (i.e the number of layers of gates in the circuit) to achieve supremacy. On the other hand, we now understand better how the particular choice of the quantum gates used to build random quantum circuits affects the simulation cost, leading to improved benchmarks for near-term quantum supremacy (available for download here), which are in some cases quadratically more expensive to simulate classically than the original proposal.

Sampling from random quantum circuits is an excellent calibration benchmark for quantum computers, which we call cross-entropy benchmarking. A successful quantum supremacy experiment with random circuits would demonstrate the basic building blocks for a large-scale fault-tolerant quantum computer. Furthermore, quantum physics has not yet been tested for highly complex quantum states such as this.
Space-time volume of a quantum circuit computation. The computational cost for quantum simulation increases with the volume of the quantum circuit, and in general grows exponentially with the number of qubits and the circuit depth. For asymmetric grids of qubits, the computational space-time volume grows slower with depth than for symmetric grids, and can result in circuits exponentially easier to simulate.
In “A blueprint for demonstrating quantum supremacy with superconducting qubits” (arXiv here), we illustrate a blueprint towards quantum supremacy and experimentally demonstrate a proof-of-principle version for the first time. In the paper, we discuss two key ingredients for quantum supremacy: exponential complexity and accurate computations. We start by running algorithms on subsections of the device ranging from 5 to 9 qubits. We find that the classical simulation cost grows exponentially with the number of qubits. These results are intended to provide a clear example of the exponential power of these devices. Next, we use cross-entropy benchmarking to compare our results against that of an ordinary computer and show that our computations are highly accurate. In fact, the error rate is low enough to achieve quantum supremacy with a larger quantum processor.

Beyond achieving quantum supremacy, a quantum platform should offer clear applications. In our paper, we apply our algorithms towards computational problems in quantum statistical-mechanics using complex multi-qubit gates (as opposed to the two-qubit gates designed for a digital quantum processor with surface code error correction). We show that our devices can be used to study fundamental properties of materials, e.g. microscopic differences between metals and insulators. By extending these results to next-generation devices with ~50 qubits, we hope to answer scientific questions that are beyond the capabilities of any other computing platform.
Photograph of two gmon superconducting qubits and their tunable coupler developed by Charles Neill and Pedram Roushan.
These two publications introduce a realistic proposal for near-term quantum supremacy, and demonstrate a proof-of-principle version for the first time. We will continue to decrease the error rates and increase the number of qubits in quantum processors to reach the quantum supremacy frontier, and to develop quantum algorithms for useful near-term applications.

The Question of Quantum Supremacy



Quantum computing integrates the two largest technological revolutions of the last half century, information technology and quantum mechanics. If we compute using the rules of quantum mechanics, instead of binary logic, some intractable computational tasks become feasible. An important goal in the pursuit of a universal quantum computer is the determination of the smallest computational task that is prohibitively hard for today’s classical computers. This crossover point is known as the “quantum supremacy” frontier, and is a critical step on the path to more powerful and useful computations.

In “Characterizing quantum supremacy in near-term devices” published in Nature Physics (arXiv here), we present the theoretical foundation for a practical demonstration of quantum supremacy in near-term devices. It proposes the task of sampling bit-strings from the output of random quantum circuits, which can be thought of as the “hello world” program for quantum computers. The upshot of the argument is that the output of random chaotic systems (think butterfly effect) become very quickly harder to predict the longer they run. If one makes a random, chaotic qubit system and examines how long a classical system would take to emulate it, one gets a good measure of when a quantum computer could outperform a classical one. Arguably, this is the strongest theoretical proposal to prove an exponential separation between the computational power of classical and quantum computers.

Determining where exactly the quantum supremacy frontier lies for sampling random quantum circuits has rapidly become an exciting area of research. On one hand, improvements in classical algorithms to simulate quantum circuits aim to increase the size of the quantum circuits required to establish quantum supremacy. This forces an experimental quantum device with a sufficiently large number of qubits and low enough error rates to implement circuits of sufficient depth (i.e the number of layers of gates in the circuit) to achieve supremacy. On the other hand, we now understand better how the particular choice of the quantum gates used to build random quantum circuits affects the simulation cost, leading to improved benchmarks for near-term quantum supremacy (available for download here), which are in some cases quadratically more expensive to simulate classically than the original proposal.

Sampling from random quantum circuits is an excellent calibration benchmark for quantum computers, which we call cross-entropy benchmarking. A successful quantum supremacy experiment with random circuits would demonstrate the basic building blocks for a large-scale fault-tolerant quantum computer. Furthermore, quantum physics has not yet been tested for highly complex quantum states such as this.
Space-time volume of a quantum circuit computation. The computational cost for quantum simulation increases with the volume of the quantum circuit, and in general grows exponentially with the number of qubits and the circuit depth. For asymmetric grids of qubits, the computational space-time volume grows slower with depth than for symmetric grids, and can result in circuits exponentially easier to simulate.
In “A blueprint for demonstrating quantum supremacy with superconducting qubits” (arXiv here), we illustrate a blueprint towards quantum supremacy and experimentally demonstrate a proof-of-principle version for the first time. In the paper, we discuss two key ingredients for quantum supremacy: exponential complexity and accurate computations. We start by running algorithms on subsections of the device ranging from 5 to 9 qubits. We find that the classical simulation cost grows exponentially with the number of qubits. These results are intended to provide a clear example of the exponential power of these devices. Next, we use cross-entropy benchmarking to compare our results against that of an ordinary computer and show that our computations are highly accurate. In fact, the error rate is low enough to achieve quantum supremacy with a larger quantum processor.

Beyond achieving quantum supremacy, a quantum platform should offer clear applications. In our paper, we apply our algorithms towards computational problems in quantum statistical-mechanics using complex multi-qubit gates (as opposed to the two-qubit gates designed for a digital quantum processor with surface code error correction). We show that our devices can be used to study fundamental properties of materials, e.g. microscopic differences between metals and insulators. By extending these results to next-generation devices with ~50 qubits, we hope to answer scientific questions that are beyond the capabilities of any other computing platform.
Photograph of two gmon superconducting qubits and their tunable coupler developed by Charles Neill and Pedram Roushan.
These two publications introduce a realistic proposal for near-term quantum supremacy, and demonstrate a proof-of-principle version for the first time. We will continue to decrease the error rates and increase the number of qubits in quantum processors to reach the quantum supremacy frontier, and to develop quantum algorithms for useful near-term applications.

Source: Google AI Blog


Google at ICLR 2018



This week, Vancouver, Canada hosts the 6th International Conference on Learning Representations (ICLR 2018), a conference focused on how one can learn meaningful and useful representations of data for machine learning. ICLR includes conference and workshop tracks, with invited talks along with oral and poster presentations of some of the latest research on deep learning, metric learning, kernel learning, compositional models, non-linear structured prediction, and issues regarding non-convex optimization.

At the forefront of innovation in cutting-edge technology in neural networks and deep learning, Google focuses on both theory and application, developing learning approaches to understand and generalize. As Platinum Sponsor of ICLR 2018, Google will have a strong presence with over 130 researchers attending, contributing to and learning from the broader academic research community by presenting papers and posters, in addition to participating on organizing committees and in workshops.

If you are attending ICLR 2018, we hope you'll stop by our booth and chat with our researchers about the projects and opportunities at Google that go into solving interesting problems for billions of people. You can also learn more about our research being presented at ICLR 2018 in the list below (Googlers highlighted in blue)

Senior Program Chairs include:
Tara Sainath

Steering Committee includes:
Hugo Larochelle

Oral Contributions
Wasserstein Auto-Encoders
Ilya Tolstikhin, Olivier Bousquet, Sylvain Gelly, Bernhard Scholkopf

On the Convergence of Adam and Beyond (Best Paper Award)
Sashank J. Reddi, Satyen Kale, Sanjiv Kumar

Ask the Right Questions: Active Question Reformulation with Reinforcement Learning
Christian Buck, Jannis Bulian, Massimiliano Ciaramita, Wojciech Gajewski, Andrea Gesmundo, Neil Houlsby, Wei Wang

Beyond Word Importance: Contextual Decompositions to Extract Interactions from LSTMs
W. James Murdoch, Peter J. Liu, Bin Yu

Conference Posters
Boosting the Actor with Dual Critic
Bo Dai, Albert Shaw, Niao He, Lihong Li, Le Song

MaskGAN: Better Text Generation via Filling in the _______
William Fedus, Ian Goodfellow, Andrew M. Dai

Scalable Private Learning with PATE
Nicolas Papernot, Shuang Song, Ilya Mironov, Ananth Raghunathan, Kunal Talwar, Ulfar Erlingsson

Deep Gradient Compression: Reducing the Communication Bandwidth for Distributed Training
Yujun Lin, Song Han, Huizi Mao, Yu Wang, William J. Dally

Flipout: Efficient Pseudo-Independent Weight Perturbations on Mini-Batches
Yeming Wen, Paul Vicol, Jimmy Ba, Dustin Tran, Roger Grosse

Latent Constraints: Learning to Generate Conditionally from Unconditional Generative Models
Adam Roberts, Jesse Engel, Matt Hoffman

Multi-Mention Learning for Reading Comprehension with Neural Cascades
Swabha Swayamdipta, Ankur P. Parikh, Tom Kwiatkowski

QANet: Combining Local Convolution with Global Self-Attention for Reading Comprehension
Adams Wei Yu, David Dohan, Thang Luong, Rui Zhao, Kai Chen, Mohammad Norouzi, Quoc V. Le

Sensitivity and Generalization in Neural Networks: An Empirical Study
Roman Novak, Yasaman Bahri, Daniel A. Abolafia, Jeffrey Pennington, Jascha Sohl-Dickstein

Action-dependent Control Variates for Policy Optimization via Stein Identity
Hao Liu, Yihao Feng, Yi Mao, Dengyong Zhou, Jian Peng, Qiang Liu

An Efficient Framework for Learning Sentence Representations
Lajanugen Logeswaran, Honglak Lee

Fidelity-Weighted Learning
Mostafa Dehghani, Arash Mehrjou, Stephan Gouws, Jaap Kamps, Bernhard Schölkopf

Generating Wikipedia by Summarizing Long Sequences
Peter J. Liu, Mohammad Saleh, Etienne Pot, Ben Goodrich, Ryan Sepassi, Lukasz Kaiser, Noam Shazeer

Matrix Capsules with EM Routing
Geoffrey Hinton, Sara Sabour, Nicholas Frosst

Temporal Difference Models: Model-Free Deep RL for Model-Based Control
Sergey Levine, Shixiang Gu, Murtaza Dalal, Vitchyr Pong

Deep Neural Networks as Gaussian Processes
Jaehoon Lee, Yasaman Bahri, Roman Novak, Samuel L. Schoenholz, Jeffrey Pennington, Jascha Sohl-Dickstein

Many Paths to Equilibrium: GANs Do Not Need to Decrease a Divergence at Every Step
William Fedus, Mihaela Rosca, Balaji Lakshminarayanan, Andrew M. Dai, Shakir Mohamed, Ian Goodfellow

Initialization Matters: Orthogonal Predictive State Recurrent Neural Networks
Krzysztof Choromanski, Carlton Downey, Byron Boots

Learning Differentially Private Recurrent Language Models
H. Brendan McMahan, Daniel Ramage, Kunal Talwar, Li Zhang

Learning Latent Permutations with Gumbel-Sinkhorn Networks
Gonzalo Mena, David Belanger, Scott Linderman, Jasper Snoek

Leave no Trace: Learning to Reset for Safe and Autonomous Reinforcement Learning
Benjamin Eysenbach, Shixiang Gu, Julian IbarzSergey Levine

Meta-Learning for Semi-Supervised Few-Shot Classification
Mengye Ren, Eleni Triantafillou, Sachin Ravi, Jake Snell, Kevin Swersky, Josh Tenenbaum, Hugo Larochelle, Richard Zemel

Thermometer Encoding: One Hot Way to Resist Adversarial Examples
Jacob Buckman, Aurko Roy, Colin Raffel, Ian Goodfellow

A Hierarchical Model for Device Placement
Azalia Mirhoseini, Anna Goldie, Hieu Pham, Benoit Steiner, Quoc V. LeJeff Dean

Monotonic Chunkwise Attention
Chung-Cheng Chiu, Colin Raffel

Training Confidence-calibrated Classifiers for Detecting Out-of-Distribution Samples
Kimin Lee, Honglak Lee, Kibok Lee, Jinwoo Shin

Trust-PCL: An Off-Policy Trust Region Method for Continuous Control
Ofir Nachum, Mohammad Norouzi, Kelvin Xu, Dale Schuurmans

Ensemble Adversarial Training: Attacks and Defenses
Florian Tramèr, Alexey Kurakin, Nicolas Papernot, Ian Goodfellow, Dan Boneh, Patrick McDaniel

Stochastic Variational Video Prediction
Mohammad Babaeizadeh, Chelsea Finn, Dumitru Erhan, Roy Campbell, Sergey Levine

Depthwise Separable Convolutions for Neural Machine Translation
Lukasz Kaiser, Aidan N. Gomez, Francois Chollet

Don’t Decay the Learning Rate, Increase the Batch Size
Samuel L. Smith, Pieter-Jan Kindermans, Chris Ying, Quoc V. Le

Generative Models of Visually Grounded Imagination
Ramakrishna Vedantam, Ian Fischer, Jonathan Huang, Kevin Murphy

Large Scale Distributed Neural Network Training through Online Distillation
Rohan Anil, Gabriel Pereyra, Alexandre Passos, Robert Ormandi, George E. Dahl, Geoffrey E. Hinton

Learning a Neural Response Metric for Retinal Prosthesis
Nishal P. Shah, Sasidhar Madugula, Alan Litke, Alexander Sher, EJ Chichilnisky, Yoram Singer, Jonathon Shlens

Neumann Optimizer: A Practical Optimization Algorithm for Deep Neural Networks
Shankar Krishnan, Ying Xiao, Rif A. Saurous

A Neural Representation of Sketch Drawings
David HaDouglas Eck

Deep Bayesian Bandits Showdown: An Empirical Comparison of Bayesian Deep Networks for Thompson Sampling
Carlos Riquelme, George Tucker, Jasper Snoek

Generalizing Hamiltonian Monte Carlo with Neural Networks
Daniel Levy, Matthew D. HoffmanJascha Sohl-Dickstein

Leveraging Grammar and Reinforcement Learning for Neural Program Synthesis
Rudy Bunel, Matthew Hausknecht, Jacob Devlin, Rishabh Singh, Pushmeet Kohli

On the Discrimination-Generalization Tradeoff in GANs
Pengchuan Zhang, Qiang Liu, Dengyong Zhou, Tao Xu, Xiaodong He

A Bayesian Perspective on Generalization and Stochastic Gradient Descent
Samuel L. Smith, Quoc V. Le

Learning how to Explain Neural Networks: PatternNet and PatternAttribution
Pieter-Jan Kindermans, Kristof T. Schütt, Maximilian Alber, Klaus-Robert Müller, Dumitru Erhan, Been Kim, Sven Dähne

Skip RNN: Learning to Skip State Updates in Recurrent Neural Networks
Víctor Campos, Brendan Jou, Xavier Giró-i-Nieto, Jordi Torres, Shih-Fu Chang

Towards Neural Phrase-based Machine Translation
Po-Sen Huang, Chong Wang, Sitao Huang, Dengyong Zhou, Li Deng

Unsupervised Cipher Cracking Using Discrete GANs
Aidan N. Gomez, Sicong Huang, Ivan Zhang, Bryan M. Li, Muhammad Osama, Lukasz Kaiser

Variational Image Compression With A Scale Hyperprior
Johannes Ballé, David Minnen, Saurabh Singh, Sung Jin Hwang, Nick Johnston

Workshop Posters
Local Explanation Methods for Deep Neural Networks Lack Sensitivity to Parameter Values
Julius Adebayo, Justin Gilmer, Ian Goodfellow, Been Kim

Stoachastic Gradient Langevin Dynamics that Exploit Neural Network Structure
Zachary Nado, Jasper Snoek, Bowen Xu, Roger Grosse, David Duvenaud, James Martens

Towards Mixed-initiative generation of multi-channel sequential structure
Anna Huang, Sherol Chen, Mark J. Nelson, Douglas Eck

Can Deep Reinforcement Learning Solve Erdos-Selfridge-Spencer Games?
Maithra Raghu, Alex Irpan, Jacob Andreas, Robert Kleinberg, Quoc V. Le, Jon Kleinberg

GILBO: One Metric to Measure Them All
Alexander Alemi, Ian Fischer

HoME: a Household Multimodal Environment
Simon Brodeur, Ethan Perez, Ankesh Anand, Florian Golemo, Luca Celotti, Florian Strub, Jean Rouat, Hugo Larochelle, Aaron Courville

Learning to Learn without Labels
Luke Metz, Niru Maheswaranathan, Brian Cheung, Jascha Sohl-Dickstein

Learning via Social Awareness: Improving Sketch Representations with Facial Feedback
Natasha Jaques, Jesse Engel, David Ha, Fred Bertsch, Rosalind Picard, Douglas Eck

Negative Eigenvalues of the Hessian in Deep Neural Networks
Guillaume Alain, Nicolas Le Roux, Pierre-Antoine Manzagol

Realistic Evaluation of Semi-Supervised Learning Algorithms
Avital Oliver, Augustus Odena, Colin Raffel, Ekin Cubuk, lan Goodfellow

Winner's Curse? On Pace, Progress, and Empirical Rigor
D. Sculley, Jasper Snoek, Alex Wiltschko, Ali Rahimi

Meta-Learning for Batch Mode Active Learning
Sachin Ravi, Hugo Larochelle

To Prune, or Not to Prune: Exploring the Efficacy of Pruning for Model Compression
Michael Zhu, Suyog Gupta

Adversarial Spheres
Justin Gilmer, Luke Metz, Fartash Faghri, Sam Schoenholz, Maithra Raghu,,Martin Wattenberg, Ian Goodfellow

Clustering Meets Implicit Generative Models
Francesco Locatello, Damien Vincent, Ilya Tolstikhin, Gunnar Ratsch, Sylvain Gelly, Bernhard Scholkopf

Decoding Decoders: Finding Optimal Representation Spaces for Unsupervised Similarity Tasks
Vitalii Zhelezniak, Dan Busbridge, April Shen, Samuel L. Smith, Nils Y. Hammerla

Learning Longer-term Dependencies in RNNs with Auxiliary Losses
Trieu Trinh, Quoc Le, Andrew Dai, Thang Luong

Graph Partition Neural Networks for Semi-Supervised Classification
Alexander Gaunt, Danny Tarlow, Marc Brockschmidt, Raquel Urtasun, Renjie Liao, Richard Zemel

Searching for Activation Functions
Prajit Ramachandran, Barret Zoph, Quoc Le

Time-Dependent Representation for Neural Event Sequence Prediction
Yang Li, Nan Du, Samy Bengio

Faster Discovery of Neural Architectures by Searching for Paths in a Large Model
Hieu Pham, Melody Guan, Barret Zoph, Quoc V. Le, Jeff Dean

Intriguing Properties of Adversarial Examples
Ekin Dogus Cubuk, Barret Zoph, Sam Schoenholz, Quoc Le

PPP-Net: Platform-aware Progressive Search for Pareto-optimal Neural Architectures
Jin-Dong Dong, An-Chieh Cheng, Da-Cheng Juan, Wei Wei, Min Sun

The Mirage of Action-Dependent Baselines in Reinforcement Learning
George Tucker, Surya Bhupatiraju, Shixiang Gu, Richard E. Turner, Zoubin Ghahramani, Sergey Levine

Learning to Organize Knowledge with N-Gram Machines
Fan Yang, Jiazhong Nie, William W. Cohen, Ni Lao

Online variance-reducing optimization
Nicolas Le Roux, Reza Babanezhad, Pierre-Antoine Manzagol

Google at ICLR 2018



This week, Vancouver, Canada hosts the 6th International Conference on Learning Representations (ICLR 2018), a conference focused on how one can learn meaningful and useful representations of data for machine learning. ICLR includes conference and workshop tracks, with invited talks along with oral and poster presentations of some of the latest research on deep learning, metric learning, kernel learning, compositional models, non-linear structured prediction, and issues regarding non-convex optimization.

At the forefront of innovation in cutting-edge technology in neural networks and deep learning, Google focuses on both theory and application, developing learning approaches to understand and generalize. As Platinum Sponsor of ICLR 2018, Google will have a strong presence with over 130 researchers attending, contributing to and learning from the broader academic research community by presenting papers and posters, in addition to participating on organizing committees and in workshops.

If you are attending ICLR 2018, we hope you'll stop by our booth and chat with our researchers about the projects and opportunities at Google that go into solving interesting problems for billions of people. You can also learn more about our research being presented at ICLR 2018 in the list below (Googlers highlighted in blue)

Senior Program Chair:
Tara Sainath

Steering Committee includes:
Hugo Larochelle

Oral Contributions
Wasserstein Auto-Encoders
Ilya Tolstikhin, Olivier Bousquet, Sylvain Gelly, Bernhard Scholkopf

On the Convergence of Adam and Beyond (Best Paper Award)
Sashank J. Reddi, Satyen Kale, Sanjiv Kumar

Ask the Right Questions: Active Question Reformulation with Reinforcement Learning
Christian Buck, Jannis Bulian, Massimiliano Ciaramita, Wojciech Gajewski, Andrea Gesmundo, Neil Houlsby, Wei Wang

Beyond Word Importance: Contextual Decompositions to Extract Interactions from LSTMs
W. James Murdoch, Peter J. Liu, Bin Yu

Conference Posters
Boosting the Actor with Dual Critic
Bo Dai, Albert Shaw, Niao He, Lihong Li, Le Song

MaskGAN: Better Text Generation via Filling in the _______
William Fedus, Ian Goodfellow, Andrew M. Dai

Scalable Private Learning with PATE
Nicolas Papernot, Shuang Song, Ilya Mironov, Ananth Raghunathan, Kunal Talwar, Ulfar Erlingsson

Deep Gradient Compression: Reducing the Communication Bandwidth for Distributed Training
Yujun Lin, Song Han, Huizi Mao, Yu Wang, William J. Dally

Flipout: Efficient Pseudo-Independent Weight Perturbations on Mini-Batches
Yeming Wen, Paul Vicol, Jimmy Ba, Dustin Tran, Roger Grosse

Latent Constraints: Learning to Generate Conditionally from Unconditional Generative Models
Adam Roberts, Jesse Engel, Matt Hoffman

Multi-Mention Learning for Reading Comprehension with Neural Cascades
Swabha Swayamdipta, Ankur P. Parikh, Tom Kwiatkowski

QANet: Combining Local Convolution with Global Self-Attention for Reading Comprehension
Adams Wei Yu, David Dohan, Thang Luong, Rui Zhao, Kai Chen, Mohammad Norouzi, Quoc V. Le

Sensitivity and Generalization in Neural Networks: An Empirical Study
Roman Novak, Yasaman Bahri, Daniel A. Abolafia, Jeffrey Pennington, Jascha Sohl-Dickstein

Action-dependent Control Variates for Policy Optimization via Stein Identity
Hao Liu, Yihao Feng, Yi Mao, Dengyong Zhou, Jian Peng, Qiang Liu

An Efficient Framework for Learning Sentence Representations
Lajanugen Logeswaran, Honglak Lee

Fidelity-Weighted Learning
Mostafa Dehghani, Arash Mehrjou, Stephan Gouws, Jaap Kamps, Bernhard Schölkopf

Generating Wikipedia by Summarizing Long Sequences
Peter J. Liu, Mohammad Saleh, Etienne Pot, Ben Goodrich, Ryan Sepassi, Lukasz Kaiser, Noam Shazeer

Matrix Capsules with EM Routing
Geoffrey Hinton, Sara Sabour, Nicholas Frosst

Temporal Difference Models: Model-Free Deep RL for Model-Based Control
Sergey Levine, Shixiang Gu, Murtaza Dalal, Vitchyr Pong

Deep Neural Networks as Gaussian Processes
Jaehoon Lee, Yasaman Bahri, Roman Novak, Samuel L. Schoenholz, Jeffrey Pennington, Jascha Sohl-Dickstein

Many Paths to Equilibrium: GANs Do Not Need to Decrease a Divergence at Every Step
William Fedus, Mihaela Rosca, Balaji Lakshminarayanan, Andrew M. Dai, Shakir Mohamed, Ian Goodfellow

Initialization Matters: Orthogonal Predictive State Recurrent Neural Networks
Krzysztof Choromanski, Carlton Downey, Byron Boots

Learning Differentially Private Recurrent Language Models
H. Brendan McMahan, Daniel Ramage, Kunal Talwar, Li Zhang

Learning Latent Permutations with Gumbel-Sinkhorn Networks
Gonzalo Mena, David Belanger, Scott Linderman, Jasper Snoek

Leave no Trace: Learning to Reset for Safe and Autonomous Reinforcement Learning
Benjamin Eysenbach, Shixiang Gu, Julian IbarzSergey Levine

Meta-Learning for Semi-Supervised Few-Shot Classification
Mengye Ren, Eleni Triantafillou, Sachin Ravi, Jake Snell, Kevin Swersky, Josh Tenenbaum, Hugo Larochelle, Richard Zemel

Thermometer Encoding: One Hot Way to Resist Adversarial Examples
Jacob Buckman, Aurko Roy, Colin Raffel, Ian Goodfellow

A Hierarchical Model for Device Placement
Azalia Mirhoseini, Anna Goldie, Hieu Pham, Benoit Steiner, Quoc V. LeJeff Dean

Monotonic Chunkwise Attention
Chung-Cheng Chiu, Colin Raffel

Training Confidence-calibrated Classifiers for Detecting Out-of-Distribution Samples
Kimin Lee, Honglak Lee, Kibok Lee, Jinwoo Shin

Trust-PCL: An Off-Policy Trust Region Method for Continuous Control
Ofir Nachum, Mohammad Norouzi, Kelvin Xu, Dale Schuurmans

Ensemble Adversarial Training: Attacks and Defenses
Florian Tramèr, Alexey Kurakin, Nicolas Papernot, Ian Goodfellow, Dan Boneh, Patrick McDaniel

Stochastic Variational Video Prediction
Mohammad Babaeizadeh, Chelsea Finn, Dumitru Erhan, Roy Campbell, Sergey Levine

Depthwise Separable Convolutions for Neural Machine Translation
Lukasz Kaiser, Aidan N. Gomez, Francois Chollet

Don’t Decay the Learning Rate, Increase the Batch Size
Samuel L. Smith, Pieter-Jan Kindermans, Chris Ying, Quoc V. Le

Generative Models of Visually Grounded Imagination
Ramakrishna Vedantam, Ian Fischer, Jonathan Huang, Kevin Murphy

Large Scale Distributed Neural Network Training through Online Distillation
Rohan Anil, Gabriel Pereyra, Alexandre Passos, Robert Ormandi, George E. Dahl, Geoffrey E. Hinton

Learning a Neural Response Metric for Retinal Prosthesis
Nishal P. Shah, Sasidhar Madugula, Alan Litke, Alexander Sher, EJ Chichilnisky, Yoram Singer, Jonathon Shlens

Neumann Optimizer: A Practical Optimization Algorithm for Deep Neural Networks
Shankar Krishnan, Ying Xiao, Rif A. Saurous

A Neural Representation of Sketch Drawings
David HaDouglas Eck

Deep Bayesian Bandits Showdown: An Empirical Comparison of Bayesian Deep Networks for Thompson Sampling
Carlos Riquelme, George Tucker, Jasper Snoek

Generalizing Hamiltonian Monte Carlo with Neural Networks
Daniel Levy, Matthew D. HoffmanJascha Sohl-Dickstein

Leveraging Grammar and Reinforcement Learning for Neural Program Synthesis
Rudy Bunel, Matthew Hausknecht, Jacob Devlin, Rishabh Singh, Pushmeet Kohli

On the Discrimination-Generalization Tradeoff in GANs
Pengchuan Zhang, Qiang Liu, Dengyong Zhou, Tao Xu, Xiaodong He

A Bayesian Perspective on Generalization and Stochastic Gradient Descent
Samuel L. Smith, Quoc V. Le

Learning how to Explain Neural Networks: PatternNet and PatternAttribution
Pieter-Jan Kindermans, Kristof T. Schütt, Maximilian Alber, Klaus-Robert Müller, Dumitru Erhan, Been Kim, Sven Dähne

Skip RNN: Learning to Skip State Updates in Recurrent Neural Networks
Víctor Campos, Brendan Jou, Xavier Giró-i-Nieto, Jordi Torres, Shih-Fu Chang

Towards Neural Phrase-based Machine Translation
Po-Sen Huang, Chong Wang, Sitao Huang, Dengyong Zhou, Li Deng

Unsupervised Cipher Cracking Using Discrete GANs
Aidan N. Gomez, Sicong Huang, Ivan Zhang, Bryan M. Li, Muhammad Osama, Lukasz Kaiser

Variational Image Compression With A Scale Hyperprior
Johannes Ballé, David Minnen, Saurabh Singh, Sung Jin Hwang, Nick Johnston

Workshop Posters
Local Explanation Methods for Deep Neural Networks Lack Sensitivity to Parameter Values
Julius Adebayo, Justin Gilmer, Ian Goodfellow, Been Kim

Stoachastic Gradient Langevin Dynamics that Exploit Neural Network Structure
Zachary Nado, Jasper Snoek, Bowen Xu, Roger Grosse, David Duvenaud, James Martens

Towards Mixed-initiative generation of multi-channel sequential structure
Anna Huang, Sherol Chen, Mark J. Nelson, Douglas Eck

Can Deep Reinforcement Learning Solve Erdos-Selfridge-Spencer Games?
Maithra Raghu, Alex Irpan, Jacob Andreas, Robert Kleinberg, Quoc V. Le, Jon Kleinberg

GILBO: One Metric to Measure Them All
Alexander Alemi, Ian Fischer

HoME: a Household Multimodal Environment
Simon Brodeur, Ethan Perez, Ankesh Anand, Florian Golemo, Luca Celotti, Florian Strub, Jean Rouat, Hugo Larochelle, Aaron Courville

Learning to Learn without Labels
Luke Metz, Niru Maheswaranathan, Brian Cheung, Jascha Sohl-Dickstein

Learning via Social Awareness: Improving Sketch Representations with Facial Feedback
Natasha Jaques, Jesse Engel, David Ha, Fred Bertsch, Rosalind Picard, Douglas Eck

Negative Eigenvalues of the Hessian in Deep Neural Networks
Guillaume Alain, Nicolas Le Roux, Pierre-Antoine Manzagol

Realistic Evaluation of Semi-Supervised Learning Algorithms
Avital Oliver, Augustus Odena, Colin Raffel, Ekin Cubuk, lan Goodfellow

Winner's Curse? On Pace, Progress, and Empirical Rigor
D. Sculley, Jasper Snoek, Alex Wiltschko, Ali Rahimi

Meta-Learning for Batch Mode Active Learning
Sachin Ravi, Hugo Larochelle

To Prune, or Not to Prune: Exploring the Efficacy of Pruning for Model Compression
Michael Zhu, Suyog Gupta

Adversarial Spheres
Justin Gilmer, Luke Metz, Fartash Faghri, Sam Schoenholz, Maithra Raghu,,Martin Wattenberg, Ian Goodfellow

Clustering Meets Implicit Generative Models
Francesco Locatello, Damien Vincent, Ilya Tolstikhin, Gunnar Ratsch, Sylvain Gelly, Bernhard Scholkopf

Decoding Decoders: Finding Optimal Representation Spaces for Unsupervised Similarity Tasks
Vitalii Zhelezniak, Dan Busbridge, April Shen, Samuel L. Smith, Nils Y. Hammerla

Learning Longer-term Dependencies in RNNs with Auxiliary Losses
Trieu Trinh, Quoc Le, Andrew Dai, Thang Luong

Graph Partition Neural Networks for Semi-Supervised Classification
Alexander Gaunt, Danny Tarlow, Marc Brockschmidt, Raquel Urtasun, Renjie Liao, Richard Zemel

Searching for Activation Functions
Prajit Ramachandran, Barret Zoph, Quoc Le

Time-Dependent Representation for Neural Event Sequence Prediction
Yang Li, Nan Du, Samy Bengio

Faster Discovery of Neural Architectures by Searching for Paths in a Large Model
Hieu Pham, Melody Guan, Barret Zoph, Quoc V. Le, Jeff Dean

Intriguing Properties of Adversarial Examples
Ekin Dogus Cubuk, Barret Zoph, Sam Schoenholz, Quoc Le

PPP-Net: Platform-aware Progressive Search for Pareto-optimal Neural Architectures
Jin-Dong Dong, An-Chieh Cheng, Da-Cheng Juan, Wei Wei, Min Sun

The Mirage of Action-Dependent Baselines in Reinforcement Learning
George Tucker, Surya Bhupatiraju, Shixiang Gu, Richard E. Turner, Zoubin Ghahramani, Sergey Levine

Learning to Organize Knowledge with N-Gram Machines
Fan Yang, Jiazhong Nie, William W. Cohen, Ni Lao

Online variance-reducing optimization
Nicolas Le Roux, Reza Babanezhad, Pierre-Antoine Manzagol

Source: Google AI Blog


Expressive Speech Synthesis with Tacotron



At Google, we're excited about the recent rapid progress of neural network-based text-to-speech (TTS) research. In particular, end-to-end architectures, such as the Tacotron systems we announced last year, can both simplify voice building pipelines and produce natural-sounding speech. This will help us build better human-computer interfaces, like conversational assistants, audiobook narration, news readers, or voice design software. To deliver a truly human-like voice, however, a TTS system must learn to model prosody, the collection of expressive factors of speech, such as intonation, stress, and rhythm. Most current end-to-end systems, including Tacotron, don't explicitly model prosody, meaning they can't control exactly how the generated speech should sound. This may lead to monotonous-sounding speech, even when models are trained on very expressive datasets like audiobooks, which often contain character voices with significant variation. Today, we are excited to share two new papers that address these problems.

Our first paper, “Towards End-to-End Prosody Transfer for Expressive Speech Synthesis with Tacotron”, introduces the concept of a prosody embedding. We augment the Tacotron architecture with an additional prosody encoder that computes a low-dimensional embedding from a clip of human speech (the reference audio).
We augment Tacotron with a prosody encoder. The lower half of the image is the original Tacotron sequence-to-sequence model. For technical details, please refer to the paper.
This embedding captures characteristics of the audio that are independent of phonetic information and idiosyncratic speaker traits — these are attributes like stress, intonation, and timing. At inference time, we can use this embedding to perform prosody transfer, generating speech in the voice of a completely different speaker, but exhibiting the prosody of the reference.

Text: *Is* that Utah travel agency?
Reference prosody (Australian)
Synthesized without prosody embedding (American)
Synthesized with prosody embedding (American)

The embedding can also transfer fine time-aligned prosody from one phrase to a slightly different phrase, though this technique works best when the reference and target phrases are similar in length and structure.

Reference Text: For the first time in her life she had been danced tired.
Synthesized Text: For the last time in his life he had been handily embarrassed.
Reference prosody (American)
Synthesized without prosody embedding (American)
Synthesized with prosody embedding (American)

Excitingly, we observe prosody transfer even when the reference audio comes from a speaker whose voice is not in Tacotron's training data.

Text: I've Swallowed a Pollywog.
Reference prosody (Unseen American Speaker)
Synthesized without prosody embedding (British)
Synthesized with prosody embedding (British)

This is a promising result, as it paves the way for voice interaction designers to use their own voice to customize speech synthesis. You can listen to the full set of audio demos for “Towards End-to-End Prosody Transfer for Expressive Speech Synthesis with Tacotron” on this web page.

Despite their ability to transfer prosody with high fidelity, the embeddings from the paper above don't completely disentangle prosody from the content of a reference audio clip. (This explains why they transfer prosody best to phrases of similar structure and length.) Furthermore, they require a clip of reference audio at inference time. A natural question then arises: can we develop a model of expressive speech that alleviates these problems?

In our second paper, “Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis”, we do just that. Building upon the architecture in our first paper, we propose a new unsupervised method for modeling latent "factors" of speech. The key to this model is that, rather than learning fine time-aligned prosodic elements, it learns higher-level speaking style patterns that can be transferred across arbitrarily different phrases.

The model works by adding an extra attention mechanism to Tacotron, forcing it to represent the prosody embedding of any speech clip as the linear combination of a fixed set of basis embeddings. We call these embeddings Global Style Tokens (GSTs), and find that they learn text-independent variations in a speaker's style (soft, high-pitch, intense, etc.), without the need for explicit style labels.
Model architecture of Global Style Tokens. The prosody embedding is decomposed into “style tokens” to enable unsupervised style control and transfer. For technical details, please refer to the paper.
At inference time, we can select or modify the combination weights for the tokens, allowing us to force Tacotron to use a specific speaking style without needing a reference audio clip. Using GSTs, for example, we can make different sentences of varying lengths sound more "lively", "angry", "lamenting", etc:

Text: United Airlines five six three from Los Angeles to New Orleans has Landed.
Style 1
Style 2
Style 3
Style 4
Style 5
The text-independent nature of GSTs make them ideal for style transfer, which takes a reference audio clip spoken in a specific style and transfers its style to any target phrase we choose. To achieve this, we first run inference to predict the GST combination weights for an utterance whose style we want to imitate. We can then feed those combination weights to the model to synthesize completely different phrases — even those with very different lengths and structure — in the same style.

Finally, our paper shows that Global Style Tokens can model more than just speaking style. When trained on noisy YouTube audio from unlabeled speakers, a GST-enabled Tacotron learns to represent noise sources and distinct speakers as separate tokens. This means that by selecting the GSTs we use in inference, we can synthesize speech free of background noise, or speech in the voice of a specific unlabeled speaker from the dataset. This exciting result provides a path towards highly scalable but robust speech synthesis. You can listen to the full set of demos for "Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis" on this web page.

We are excited about the potential applications and opportunities that these two bodies of research enable. In the meantime, there are new important research problems to be addressed. We'd like to extend the techniques of the first paper to support prosody transfer in the natural pitch range of the target speaker. We'd also like to develop techniques to select appropriate prosody or speaking style automatically from context, using, for example, the integration of natural language understanding with TTS. Finally, while our first paper proposes an initial set of objective and subjective metrics for prosody transfer, we'd like to develop these further to help establish generally-accepted methods for prosodic evaluation.

Acknowledgements
These projects were done jointly between multiple Google teams. Contributors include RJ Skerry-Ryan, Yuxuan Wang, Daisy Stanton, Eric Battenberg, Ying Xiao, Joel Shor, Rif A. Saurous, Yu Zhang, Ron J. Weiss, Rob Clark, Fei Ren and Ye Jia.


Expressive Speech Synthesis with Tacotron

At Google, we’re excited about the recent rapid progress of neural network-based text-to-speech (TTS) research. In particular, end-to-end architectures, such as the Tacotron systems we announced last year, can both simplify voice building pipelines and produce natural-sounding speech. This will help us build better human-computer interfaces, like conversational assistants, audiobook narration, news readers, or voice design software. To deliver a truly human-like voice, however, a TTS system must learn to model prosody, the collection of expressive factors of speech, such as intonation, stress, and rhythm. Most current end-to-end systems, including Tacotron, don’t explicitly model prosody, meaning they can’t control exactly how the generated speech should sound. This may lead to monotonous-sounding speech, even when models are trained on very expressive datasets like audiobooks, which often contain character voices with significant variation. Today, we are excited to share two new papers that address these problems.

Our first paper, “Towards End-to-End Prosody Transfer for Expressive Speech Synthesis with Tacotron”, introduces the concept of a prosody embedding. We augment the Tacotron architecture with an additional prosody encoder that computes a low-dimensional embedding from a clip of human speech (the reference audio).

We augment Tacotron with a prosody encoder. The lower half of the image is the original Tacotron sequence-to-sequence model. For technical details, please refer to the paper.

This embedding captures characteristics of the audio that are independent of phonetic information and idiosyncratic speaker traits — these are attributes like stress, intonation, and timing. At inference time, we can use this embedding to perform prosody transfer, generating speech in the voice of a completely different speaker, but exhibiting the prosody of the reference.

Text: *Is* that Utah travel agency?
Reference prosody (Australian)
Synthesized without prosody embedding (American)
Synthesized with prosody embedding (American)

The embedding can also transfer fine time-aligned prosody from one phrase to a slightly different phrase, though this technique works best when the reference and target phrases are similar in length and structure.

Reference Text: For the first time in her life she had been danced tired.
Synthesized Text: For the last time in his life he had been handily embarrassed.
Reference prosody (American)
Synthesized without prosody embedding (American)
Synthesized with prosody embedding (American)

Excitingly, we observe prosody transfer even when the reference audio comes from a speaker whose voice is not in Tacotron’s training data.

Text: I’ve Swallowed a Pollywog.
Reference prosody (Unseen American Speaker)
Synthesized without prosody embedding (British)
Synthesized with prosody embedding (British)

This is a promising result, as it paves the way for voice interaction designers to use their own voice to customize speech synthesis. You can listen to the full set of audio demos for “Towards End-to-End Prosody Transfer for Expressive Speech Synthesis with Tacotron” on this web page.

Despite their ability to transfer prosody with high fidelity, the embeddings from the paper above don’t completely disentangle prosody from the content of a reference audio clip. (This explains why they transfer prosody best to phrases of similar structure and length.) Furthermore, they require a clip of reference audio at inference time. A natural question then arises: can we develop a model of expressive speech that alleviates these problems?

In our second paper, “Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis”, we do just that. Building upon the architecture in our first paper, we propose a new unsupervised method for modeling latent “factors” of speech. The key to this model is that, rather than learning fine time-aligned prosodic elements, it learns higher-level speaking style patterns that can be transferred across arbitrarily different phrases.

The model works by adding an extra attention mechanism to Tacotron, forcing it to represent the prosody embedding of any speech clip as the linear combination of a fixed set of basis embeddings. We call these embeddings Global Style Tokens (GSTs), and find that they learn text-independent variations in a speaker’s style (soft, high-pitch, intense, etc.), without the need for explicit style labels.

Model architecture of Global Style Tokens. The prosody embedding is decomposed into “style tokens” to enable unsupervised style control and transfer. For technical details, please refer to the paper.

At inference time, we can select or modify the combination weights for the tokens, allowing us to force Tacotron to use a specific speaking style without needing a reference audio clip. Using GSTs, for example, we can make different sentences of varying lengths sound more “lively”, “angry”, “lamenting”, etc:

Text: United Airlines five six three from Los Angeles to New Orleans has Landed.
Style 1
Style 2
Style 3
Style 4
Style 5

The text-independent nature of GSTs make them ideal for style transfer, which takes a reference audio clip spoken in a specific style and transfers its style to any target phrase we choose. To achieve this, we first run inference to predict the GST combination weights for an utterance whose style we want to imitate. We can then feed those combination weights to the model to synthesize completely different phrases — even those with very different lengths and structure — in the same style.

Finally, our paper shows that Global Style Tokens can model more than just speaking style. When trained on noisy YouTube audio from unlabeled speakers, a GST-enabled Tacotron learns to represent noise sources and distinct speakers as separate tokens. This means that by selecting the GSTs we use in inference, we can synthesize speech free of background noise, or speech in the voice of a specific unlabeled speaker from the dataset. This exciting result provides a path towards highly scalable but robust speech synthesis. You can listen to the full set of demos for “Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis” on this web page.

We are excited about the potential applications and opportunities that these two bodies of research enable. In the meantime, there are new important research problems to be addressed. We’d like to extend the techniques of the first paper to support prosody transfer in the natural pitch range of the target speaker. We’d also like to develop techniques to select appropriate prosody or speaking style automatically from context, using, for example, the integration of natural language understanding with TTS. Finally, while our first paper proposes an initial set of objective and subjective metrics for prosody transfer, we’d like to develop these further to help establish generally-accepted methods for prosodic evaluation.

Acknowledgements
These projects were done jointly between multiple Google teams. Contributors include RJ Skerry-Ryan, Yuxuan Wang, Daisy Stanton, Eric Battenberg, Ying Xiao, Joel Shor, Rif A. Saurous, Yu Zhang, Ron J. Weiss, Rob Clark, Fei Ren and Ye Jia.

Using Evolutionary AutoML to Discover Neural Network Architectures



The brain has evolved over a long time, from very simple worm brains 500 million years ago to a diversity of modern structures today. The human brain, for example, can accomplish a wide variety of activities, many of them effortlessly — telling whether a visual scene contains animals or buildings feels trivial to us, for example. To perform activities like these, artificial neural networks require careful design by experts over years of difficult research, and typically address one specific task, such as to find what's in a photograph, to call a genetic variant, or to help diagnose a disease. Ideally, one would want to have an automated method to generate the right architecture for any given task.

One approach to generate these architectures is through the use of evolutionary algorithms. Traditional research into neuro-evolution of topologies (e.g. Stanley and Miikkulainen 2002) has laid the foundations that allow us to apply these algorithms at scale today, and many groups are working on the subject, including OpenAI, Uber Labs, Sentient Labs and DeepMind. Of course, the Google Brain team has been thinking about AutoML too. In addition to learning-based approaches (eg. reinforcement learning), we wondered if we could use our computational resources to programmatically evolve image classifiers at unprecedented scale. Can we achieve solutions with minimal expert participation? How good can today's artificially-evolved neural networks be? We address these questions through two papers.

In “Large-Scale Evolution of Image Classifiers,” presented at ICML 2017, we set up an evolutionary process with simple building blocks and trivial initial conditions. The idea was to "sit back" and let evolution at scale do the work of constructing the architecture. Starting from very simple networks, the process found classifiers comparable to hand-designed models at the time. This was encouraging because many applications may require little user participation. For example, some users may need a better model but may not have the time to become machine learning experts. A natural question to consider next was whether a combination of hand-design and evolution could do better than either approach alone. Thus, in our more recent paper, “Regularized Evolution for Image Classifier Architecture Search” (2018), we participated in the process by providing sophisticated building blocks and good initial conditions (discussed below). Moreover, we scaled up computation using Google's new TPUv2 chips. This combination of modern hardware, expert knowledge, and evolution worked together to produce state-of-the-art models on CIFAR-10 and ImageNet, two popular benchmarks for image classification.

A Simple Approach
The following is an example of an experiment from our first paper. In the figure below, each dot is a neural network trained on the CIFAR-10 dataset, which is commonly used to train image classifiers. Initially, the population consists of one thousand identical simple seed models (no hidden layers). Starting from simple seed models is important — if we had started from a high-quality model with initial conditions containing expert knowledge, it would have been easier to get a high-quality model in the end. Once seeded with the simple models, the process advances in steps. At each step, a pair of neural networks is chosen at random. The network with higher accuracy is selected as a parent and is copied and mutated to generate a child that is then added to the population, while the other neural network dies out. All other networks remain unchanged during the step. With the application of many such steps in succession, the population evolves.
Progress of an evolution experiment. Each dot represents an individual in the population. The four diagrams show examples of discovered architectures. These correspond to the best individual (rightmost; selected by validation accuracy) and three of its ancestors.
The mutations in our first paper are purposefully simple: remove a convolution at random, add a skip connection between arbitrary layers, or change the learning rate, to name a few. This way, the results show the potential of the evolutionary algorithm, as opposed to the quality of the search space. For example, if we had used a single mutation that transforms one of the seed networks into an Inception-ResNet classifier in one step, we would be incorrectly concluding that the algorithm found a good answer. Yet, in that case, all we would have done is hard-coded the final answer into a complex mutation, rigging the outcome. If instead we stick with simple mutations, this cannot happen and evolution is truly doing the job. In the experiment in the figure, simple mutations and the selection process cause the networks to improve over time and reach high test accuracies, even though the test set had never been seen during the process. In this paper, the networks can also inherit their parent's weights. Thus, in addition to evolving the architecture, the population trains its networks while exploring the search space of initial conditions and learning-rate schedules. As a result, the process yields fully trained models with optimized hyperparameters. No expert input is needed after the experiment starts.

In all the above, even though we were minimizing the researcher's participation by having simple initial architectures and intuitive mutations, a good amount of expert knowledge went into the building blocks those architectures were made of. These included important inventions such as convolutions, ReLUs and batch-normalization layers. We were evolving an architecture made up of these components. The term "architecture" is not accidental: this is analogous to constructing a house with high-quality bricks.

Combining Evolution and Hand Design
After our first paper, we wanted to reduce the search space to something more manageable by giving the algorithm fewer choices to explore. Using our architectural analogy, we removed all the possible ways of making large-scale errors, such as putting the wall above the roof, from the search space. Similarly with neural network architecture searches, by fixing the large-scale structure of the network, we can help the algorithm out. So how to do this? The inception-like modules introduced in Zoph et al. (2017) for the purpose of architecture search proved very powerful. Their idea is to have a deep stack of repeated modules called cells. The stack is fixed but the architecture of the individual modules can change.
The building blocks introduced in Zoph et al. (2017). The diagram on the left is the outer structure of the full neural network, which parses the input data from bottom to top through a stack of repeated cells. The diagram on the right is the inside structure of a cell. The goal is to find a cell that yields an accurate network.
In our second paper, “Regularized Evolution for Image Classifier Architecture Search” (2018), we presented the results of applying evolutionary algorithms to the search space described above. The mutations modify the cell by randomly reconnecting the inputs (the arrows on the right diagram in the figure) or randomly replacing the operations (for example, they can replace the "max 3x3" in the figure, a max-pool operation, with an arbitrary alternative). These mutations are still relatively simple, but the initial conditions are not: the population is now initialized with models that must conform to the outer stack of cells, which was designed by an expert. Even though the cells in these seed models are random, we are no longer starting from simple models, which makes it easier to get to high-quality models in the end. If the evolutionary algorithm is contributing meaningfully, the final networks should be significantly better than the networks we already know can be constructed within this search space. Our paper shows that evolution can indeed find state-of-the-art models that either match or outperform hand-designs.

A Controlled Comparison
Even though the mutation/selection evolutionary process is not complicated, maybe an even more straightforward approach (like random search) could have done the same. Other alternatives, though not simpler, also exist in the literature (like reinforcement learning). Because of this, the main purpose of our second paper was to provide a controlled comparison between techniques.
Comparison between evolution, reinforcement learning, and random search for the purposes of architecture search. These experiments were done on the CIFAR-10 dataset, under the same conditions as Zoph et al. (2017), where the search space was originally used with reinforcement learning.
The figure above compares evolution, reinforcement learning, and random search. On the left, each curve represents the progress of an experiment, showing that evolution is faster than reinforcement learning in the earlier stages of the search. This is significant because with less compute power available, the experiments may have to stop early. Moreover evolution is quite robust to changes in the dataset or search space. Overall, the goal of this controlled comparison is to provide the research community with the results of a computationally expensive experiment. In doing so, it is our hope to facilitate architecture searches for everyone by providing a case study of the relationship between the different search algorithms. Note, for example, that the figure above shows that the final models obtained with evolution can reach very high accuracy while using fewer floating-point operations.

One important feature of the evolutionary algorithm we used in our second paper is a form of regularization: instead of letting the worst neural networks die, we remove the oldest ones — regardless of how good they are. This improves robustness to changes in the task being optimized and tends to produce more accurate networks in the end. One reason for this may be that since we didn't allow weight inheritance, all networks must train from scratch. Therefore, this form of regularization selects for networks that remain good when they are re-trained. In other words, because a model can be more accurate just by chance — noise in the training process means even identical architectures may get different accuracy values — only architectures that remain accurate through the generations will survive in the long run, leading to the selection of networks that retrain well. More details of this conjecture can be found in the paper.

The state-of-the-art models we evolved are nicknamed AmoebaNets, and are one of the latest results from our AutoML efforts. All these experiments took a lot of computation — we used hundreds of GPUs/TPUs for days. Much like a single modern computer can outperform thousands of decades-old machines, we hope that in the future these experiments will become household. Here we aimed to provide a glimpse into that future.

Acknowledgements
We would like to thank Alok Aggarwal, Yanping Huang, Andrew Selle, Sherry Moore, Saurabh Saxena, Yutaka Leon Suematsu, Jie Tan, Alex Kurakin, Quoc Le, Barret Zoph, Jon Shlens, Vijay Vasudevan, Vincent Vanhoucke, Megan Kacholia, Jeff Dean, and the rest of the Google Brain team for the collaborations that made this work possible.

Source: Google AI Blog


Using Evolutionary AutoML to Discover Neural Network Architectures



The brain has evolved over a long time, from very simple worm brains 500 million years ago to a diversity of modern structures today. The human brain, for example, can accomplish a wide variety of activities, many of them effortlessly — telling whether a visual scene contains animals or buildings feels trivial to us, for example. To perform activities like these, artificial neural networks require careful design by experts over years of difficult research, and typically address one specific task, such as to find what's in a photograph, to call a genetic variant, or to help diagnose a disease. Ideally, one would want to have an automated method to generate the right architecture for any given task.

One approach to generate these architectures is through the use of evolutionary algorithms. Traditional research into neuro-evolution of topologies (e.g. Stanley and Miikkulainen 2002) has laid the foundations that allow us to apply these algorithms at scale today, and many groups are working on the subject, including OpenAI, Uber Labs, Sentient Labs and DeepMind. Of course, the Google Brain team has been thinking about AutoML too. In addition to learning-based approaches (eg. reinforcement learning), we wondered if we could use our computational resources to programmatically evolve image classifiers at unprecedented scale. Can we achieve solutions with minimal expert participation? How good can today's artificially-evolved neural networks be? We address these questions through two papers.

In “Large-Scale Evolution of Image Classifiers,” presented at ICML 2017, we set up an evolutionary process with simple building blocks and trivial initial conditions. The idea was to "sit back" and let evolution at scale do the work of constructing the architecture. Starting from very simple networks, the process found classifiers comparable to hand-designed models at the time. This was encouraging because many applications may require little user participation. For example, some users may need a better model but may not have the time to become machine learning experts. A natural question to consider next was whether a combination of hand-design and evolution could do better than either approach alone. Thus, in our more recent paper, “Regularized Evolution for Image Classifier Architecture Search” (2018), we participated in the process by providing sophisticated building blocks and good initial conditions (discussed below). Moreover, we scaled up computation using Google's new TPUv2 chips. This combination of modern hardware, expert knowledge, and evolution worked together to produce state-of-the-art models on CIFAR-10 and ImageNet, two popular benchmarks for image classification.

A Simple Approach
The following is an example of an experiment from our first paper. In the figure below, each dot is a neural network trained on the CIFAR-10 dataset, which is commonly used to train image classifiers. Initially, the population consists of one thousand identical simple seed models (no hidden layers). Starting from simple seed models is important — if we had started from a high-quality model with initial conditions containing expert knowledge, it would have been easier to get a high-quality model in the end. Once seeded with the simple models, the process advances in steps. At each step, a pair of neural networks is chosen at random. The network with higher accuracy is selected as a parent and is copied and mutated to generate a child that is then added to the population, while the other neural network dies out. All other networks remain unchanged during the step. With the application of many such steps in succession, the population evolves.
Progress of an evolution experiment. Each dot represents an individual in the population. The four diagrams show examples of discovered architectures. These correspond to the best individual (rightmost; selected by validation accuracy) and three of its ancestors.
The mutations in our first paper are purposefully simple: remove a convolution at random, add a skip connection between arbitrary layers, or change the learning rate, to name a few. This way, the results show the potential of the evolutionary algorithm, as opposed to the quality of the search space. For example, if we had used a single mutation that transforms one of the seed networks into an Inception-ResNet classifier in one step, we would be incorrectly concluding that the algorithm found a good answer. Yet, in that case, all we would have done is hard-coded the final answer into a complex mutation, rigging the outcome. If instead we stick with simple mutations, this cannot happen and evolution is truly doing the job. In the experiment in the figure, simple mutations and the selection process cause the networks to improve over time and reach high test accuracies, even though the test set had never been seen during the process. In this paper, the networks can also inherit their parent's weights. Thus, in addition to evolving the architecture, the population trains its networks while exploring the search space of initial conditions and learning-rate schedules. As a result, the process yields fully trained models with optimized hyperparameters. No expert input is needed after the experiment starts.

In all the above, even though we were minimizing the researcher's participation by having simple initial architectures and intuitive mutations, a good amount of expert knowledge went into the building blocks those architectures were made of. These included important inventions such as convolutions, ReLUs and batch-normalization layers. We were evolving an architecture made up of these components. The term "architecture" is not accidental: this is analogous to constructing a house with high-quality bricks.

Combining Evolution and Hand Design
After our first paper, we wanted to reduce the search space to something more manageable by giving the algorithm fewer choices to explore. Using our architectural analogy, we removed all the possible ways of making large-scale errors, such as putting the wall above the roof, from the search space. Similarly with neural network architecture searches, by fixing the large-scale structure of the network, we can help the algorithm out. So how to do this? The inception-like modules introduced in Zoph et al. (2017) for the purpose of architecture search proved very powerful. Their idea is to have a deep stack of repeated modules called cells. The stack is fixed but the architecture of the individual modules can change.
The building blocks introduced in Zoph et al. (2017). The diagram on the left is the outer structure of the full neural network, which parses the input data from bottom to top through a stack of repeated cells. The diagram on the right is the inside structure of a cell. The goal is to find a cell that yields an accurate network.
In our second paper, “Regularized Evolution for Image Classifier Architecture Search” (2018), we presented the results of applying evolutionary algorithms to the search space described above. The mutations modify the cell by randomly reconnecting the inputs (the arrows on the right diagram in the figure) or randomly replacing the operations (for example, they can replace the "max 3x3" in the figure, a max-pool operation, with an arbitrary alternative). These mutations are still relatively simple, but the initial conditions are not: the population is now initialized with models that must conform to the outer stack of cells, which was designed by an expert. Even though the cells in these seed models are random, we are no longer starting from simple models, which makes it easier to get to high-quality models in the end. If the evolutionary algorithm is contributing meaningfully, the final networks should be significantly better than the networks we already know can be constructed within this search space. Our paper shows that evolution can indeed find state-of-the-art models that either match or outperform hand-designs.

A Controlled Comparison
Even though the mutation/selection evolutionary process is not complicated, maybe an even more straightforward approach (like random search) could have done the same. Other alternatives, though not simpler, also exist in the literature (like reinforcement learning). Because of this, the main purpose of our second paper was to provide a controlled comparison between techniques.
Comparison between evolution, reinforcement learning, and random search for the purposes of architecture search. These experiments were done on the CIFAR-10 dataset, under the same conditions as Zoph et al. (2017), where the search space was originally used with reinforcement learning.
The figure above compares evolution, reinforcement learning, and random search. On the left, each curve represents the progress of an experiment, showing that evolution is faster than reinforcement learning in the earlier stages of the search. This is significant because with less compute power available, the experiments may have to stop early. Moreover evolution is quite robust to changes in the dataset or search space. Overall, the goal of this controlled comparison is to provide the research community with the results of a computationally expensive experiment. In doing so, it is our hope to facilitate architecture searches for everyone by providing a case study of the relationship between the different search algorithms. Note, for example, that the figure above shows that the final models obtained with evolution can reach very high accuracy while using fewer floating-point operations.

One important feature of the evolutionary algorithm we used in our second paper is a form of regularization: instead of letting the worst neural networks die, we remove the oldest ones — regardless of how good they are. This improves robustness to changes in the task being optimized and tends to produce more accurate networks in the end. One reason for this may be that since we didn't allow weight inheritance, all networks must train from scratch. Therefore, this form of regularization selects for networks that remain good when they are re-trained. In other words, because a model can be more accurate just by chance — noise in the training process means even identical architectures may get different accuracy values — only architectures that remain accurate through the generations will survive in the long run, leading to the selection of networks that retrain well. More details of this conjecture can be found in the paper.

The state-of-the-art models we evolved are nicknamed AmoebaNets, and are one of the latest results from our AutoML efforts. All these experiments took a lot of computation — we used hundreds of GPUs/TPUs for days. Much like a single modern computer can outperform thousands of decades-old machines, we hope that in the future these experiments will become household. Here we aimed to provide a glimpse into that future.

Acknowledgements
We would like to thank Alok Aggarwal, Yanping Huang, Andrew Selle, Sherry Moore, Saurabh Saxena, Yutaka Leon Suematsu, Jie Tan, Alex Kurakin, Quoc Le, Barret Zoph, Jon Shlens, Vijay Vasudevan, Vincent Vanhoucke, Megan Kacholia, Jeff Dean, and the rest of the Google Brain team for the collaborations that made this work possible.

Tacotron 2: Generating Human-like Speech from Text



Generating very natural sounding speech from text (text-to-speech, TTS) has been a research goal for decades. There has been great progress in TTS research over the last few years and many individual pieces of a complete TTS system have greatly improved. Incorporating ideas from past work such as Tacotron and WaveNet, we added more improvements to end up with our new system, Tacotron 2. Our approach does not use complex linguistic and acoustic features as input. Instead, we generate human-like speech from text using neural networks trained using only speech examples and corresponding text transcripts.

A full description of our new system can be found in our paper “Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions.” In a nutshell it works like this: We use a sequence-to-sequence model optimized for TTS to map a sequence of letters to a sequence of features that encode the audio. These features, an 80-dimensional audio spectrogram with frames computed every 12.5 milliseconds, capture not only pronunciation of words, but also various subtleties of human speech, including volume, speed and intonation. Finally these features are converted to a 24 kHz waveform using a WaveNet-like architecture.
A detailed look at Tacotron 2's model architecture. The lower half of the image describes the sequence-to-sequence model that maps a sequence of letters to a spectrogram. For technical details, please refer to the paper.
You can listen to some of the Tacotron 2 audio samples that demonstrate the results of our state-of-the-art TTS system. In an evaluation where we asked human listeners to rate the naturalness of the generated speech, we obtained a score that was comparable to that of professional recordings.

While our samples sound great, there are still some difficult problems to be tackled. For example, our system has difficulties pronouncing complex words (such as “decorum” and “merlot”), and in extreme cases it can even randomly generate strange noises. Also, our system cannot yet generate audio in realtime. Furthermore, we cannot yet control the generated speech, such as directing it to sound happy or sad. Each of these is an interesting research problem on its own.

Acknowledgements
Jonathan Shen, Ruoming Pang, Ron J. Weiss, Mike Schuster, Navdeep Jaitly, Zongheng Yang, Zhifeng Chen, Yu Zhang, Yuxuan Wang, RJ Skerry-Ryan, Rif A. Saurous, Yannis Agiomyrgiannakis, Yonghui Wu, Sound Understanding team, TTS Research team, and TensorFlow team.

Google at NIPS 2017



This week, Long Beach, California hosts the 31st annual Conference on Neural Information Processing Systems (NIPS 2017), a machine learning and computational neuroscience conference that includes invited talks, demonstrations and presentations of some of the latest in machine learning research. Google will have a strong presence at NIPS 2017, with over 450 Googlers attending to contribute to, and learn from, the broader academic research community via technical talks and posters, workshops, competitions and tutorials.

Google is at the forefront of machine learning, actively exploring virtually all aspects of the field from classical algorithms to deep learning and more. Focusing on both theory and application, much of our work on language understanding, speech, translation, visual processing, and prediction relies on state-of-the-art techniques that push the boundaries of what is possible. In all of those tasks and many others, we develop learning approaches to understand and generalize, providing us with new ways of looking at old problems and helping transform how we work and live.

If you are attending NIPS 2017, we hope you’ll stop by our booth and chat with our researchers about the projects and opportunities at Google that go into solving interesting problems for billions of people, and to see demonstrations of some of the exciting research we pursue. You can also learn more about our work being presented in the list below (Googlers highlighted in blue).

Google is a Platinum Sponsor of NIPS 2017.

Organizing Committee
Program Chair: Samy Bengio
Senior Area Chairs include: Corinna Cortes, Dale Schuurmans, Hugo Larochelle
Area Chairs include: Afshin Rostamizadeh, Amir Globerson, Been Kim, D. Sculley, Dumitru Erhan, Gal Chechik, Hartmut Neven, Honglak Lee, Ian Goodfellow, Jasper Snoek, John Wright, Jon Shlens, Kun Zhang, Lihong Li, Maya Gupta, Moritz Hardt, Navdeep Jaitly, Ryan Adams, Sally Goldman, Sanjiv Kumar, Surya Ganguli, Tara Sainath, Umar Syed, Viren Jain, Vitaly Kuznetsov

Invited Talk
Powering the next 100 years
John Platt

Accepted Papers
A Meta-Learning Perspective on Cold-Start Recommendations for Items
Manasi Vartak, Hugo Larochelle, Arvind Thiagarajan

AdaGAN: Boosting Generative Models
Ilya Tolstikhin, Sylvain Gelly, Olivier Bousquet, Carl-Johann Simon-Gabriel, Bernhard Schölkopf

Deep Lattice Networks and Partial Monotonic Functions
Seungil You, David Ding, Kevin Canini, Jan Pfeifer, Maya Gupta

From which world is your graph
Cheng Li, Varun Kanade, Felix MF Wong, Zhenming Liu

Hiding Images in Plain Sight: Deep Steganography
Shumeet Baluja

Improved Graph Laplacian via Geometric Self-Consistency
Dominique Joncas, Marina Meila, James McQueen

Model-Powered Conditional Independence Test
Rajat Sen, Ananda Theertha Suresh, Karthikeyan Shanmugam, Alexandros Dimakis, Sanjay Shakkottai

Nonlinear random matrix theory for deep learning
Jeffrey Pennington, Pratik Worah

Resurrecting the sigmoid in deep learning through dynamical isometry: theory and practice
Jeffrey Pennington, Samuel Schoenholz, Surya Ganguli

SGD Learns the Conjugate Kernel Class of the Network
Amit Daniely

SVCCA: Singular Vector Canonical Correlation Analysis for Deep Learning Dynamics and Interpretability
Maithra Raghu, Justin Gilmer, Jason Yosinski, Jascha Sohl-Dickstein

Learning Hierarchical Information Flow with Recurrent Neural Modules
Danijar Hafner, Alexander Irpan, James Davidson, Nicolas Heess

Online Learning with Transductive Regret
Scott Yang, Mehryar Mohri

Acceleration and Averaging in Stochastic Descent Dynamics
Walid Krichene, Peter Bartlett

Parameter-Free Online Learning via Model Selection
Dylan J Foster, Satyen Kale, Mehryar Mohri, Karthik Sridharan

Dynamic Routing Between Capsules
Sara Sabour, Nicholas Frosst, Geoffrey E Hinton

Modulating early visual processing by language
Harm de Vries, Florian Strub, Jeremie Mary, Hugo Larochelle, Olivier Pietquin, Aaron C Courville

MarrNet: 3D Shape Reconstruction via 2.5D Sketches
Jiajun Wu, Yifan Wang, Tianfan Xue, Xingyuan Sun, Bill Freeman, Josh Tenenbaum

Affinity Clustering: Hierarchical Clustering at Scale
Mahsa Derakhshan, Soheil Behnezhad, Mohammadhossein Bateni, Vahab Mirrokni, MohammadTaghi Hajiaghayi, Silvio Lattanzi, Raimondas Kiveris

Asynchronous Parallel Coordinate Minimization for MAP Inference
Ofer Meshi, Alexander Schwing

Cold-Start Reinforcement Learning with Softmax Policy Gradient
Nan Ding, Radu Soricut

Filtering Variational Objectives
Chris J Maddison, Dieterich Lawson, George Tucker, Mohammad Norouzi, Nicolas Heess, Andriy Mnih, Yee Whye Teh, Arnaud Doucet

Multi-Armed Bandits with Metric Movement Costs
Tomer Koren, Roi Livni, Yishay Mansour

Multiscale Quantization for Fast Similarity Search
Xiang Wu, Ruiqi Guo, Ananda Theertha Suresh, Sanjiv Kumar, Daniel Holtmann-Rice, David Simcha, Felix Yu

Reducing Reparameterization Gradient Variance
Andrew Miller, Nicholas Foti, Alexander D'Amour, Ryan Adams

Statistical Cost Sharing
Eric Balkanski, Umar Syed, Sergei Vassilvitskii

The Unreasonable Effectiveness of Structured Random Orthogonal Embeddings
Krzysztof Choromanski, Mark Rowland, Adrian Weller

Value Prediction Network
Junhyuk Oh, Satinder Singh, Honglak Lee

REBAR: Low-variance, unbiased gradient estimates for discrete latent variable models
George Tucker, Andriy Mnih, Chris J Maddison, Dieterich Lawson, Jascha Sohl-Dickstein

Approximation and Convergence Properties of Generative Adversarial Learning
Shuang Liu, Olivier Bousquet, Kamalika Chaudhuri

Attention is All you Need
Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, Illia Polosukhin

PASS-GLM: polynomial approximate sufficient statistics for scalable Bayesian GLM inference
Jonathan Huggins, Ryan Adams, Tamara Broderick

Repeated Inverse Reinforcement Learning
Kareem Amin, Nan Jiang, Satinder Singh

Fair Clustering Through Fairlets
Flavio Chierichetti, Ravi Kumar, Silvio Lattanzi, Sergei Vassilvitskii

Affine-Invariant Online Optimization and the Low-rank Experts Problem
Tomer Koren, Roi Livni

Batch Renormalization: Towards Reducing Minibatch Dependence in Batch-Normalized Models
Sergey Ioffe

Bridging the Gap Between Value and Policy Based Reinforcement Learning
Ofir Nachum, Mohammad Norouzi, Kelvin Xu, Dale Schuurmans

Discriminative State Space Models
Vitaly Kuznetsov, Mehryar Mohri

Dynamic Revenue Sharing
Santiago Balseiro, Max Lin, Vahab Mirrokni, Renato Leme, Song Zuo

Multi-view Matrix Factorization for Linear Dynamical System Estimation
Mahdi Karami, Martha White, Dale Schuurmans, Csaba Szepesvari

On Blackbox Backpropagation and Jacobian Sensing
Krzysztof Choromanski, Vikas Sindhwani

On the Consistency of Quick Shift
Heinrich Jiang

Revenue Optimization with Approximate Bid Predictions
Andres Munoz, Sergei Vassilvitskii

Shape and Material from Sound
Zhoutong Zhang, Qiujia Li, Zhengjia Huang, Jiajun Wu, Josh Tenenbaum, Bill Freeman

Learning to See Physics via Visual De-animation
Jiajun Wu, Erika Lu, Pushmeet Kohli, Bill Freeman, Josh Tenenbaum

Conference Demos
Electronic Screen Protector with Efficient and Robust Mobile Vision
Hee Jung Ryu, Florian Schroff

Magenta and deeplearn.js: Real-time Control of DeepGenerative Music Models in the Browser
Curtis Hawthorne, Ian Simon, Adam Roberts, Jesse Engel, Daniel Smilkov, Nikhil Thorat, Douglas Eck

Workshops
6th Workshop on Automated Knowledge Base Construction (AKBC) 2017
Program Committee includes: Arvind Neelakanta
Authors include: Jiazhong Nie, Ni Lao

Acting and Interacting in the Real World: Challenges in Robot Learning
Invited Speakers include: Pierre Sermanet

Advances in Approximate Bayesian Inference
Panel moderator: Matthew D. Hoffman

Conversational AI - Today's Practice and Tomorrow's Potential
Invited Speakers include: Matthew Henderson, Dilek Hakkani-Tur
Organizers include: Larry Heck

Extreme Classification: Multi-class and Multi-label Learning in Extremely Large Label Spaces
Invited Speakers include: Ed Chi, Mehryar Mohri

Learning in the Presence of Strategic Behavior
Invited Speakers include: Mehryar Mohri
Presenters include: Andres Munoz Medina, Sebastien Lahaie, Sergei Vassilvitskii, Balasubramanian Sivan

Learning on Distributions, Functions, Graphs and Groups
Invited speakers include: Corinna Cortes

Machine Deception
Organizers include: Ian Goodfellow
Invited Speakers include: Jacob Buckman, Aurko Roy, Colin Raffel, Ian Goodfellow

Machine Learning and Computer Security
Invited Speakers include: Ian Goodfellow
Organizers include: Nicolas Papernot
Authors include: Jacob Buckman, Aurko Roy, Colin Raffel, Ian Goodfellow

Machine Learning for Creativity and Design
Keynote Speakers include: Ian Goodfellow
Organizers include: Doug Eck, David Ha

Machine Learning for Audio Signal Processing (ML4Audio)
Authors include: Aren Jansen, Manoj Plakal, Dan Ellis, Shawn Hershey, Channing Moore, Rif A. Saurous, Yuxuan Wang, RJ Skerry-Ryan, Ying Xiao, Daisy Stanton, Joel Shor, Eric Batternberg, Rob Clark

Machine Learning for Health (ML4H)
Organizers include: Jasper Snoek, Alex Wiltschko
Keynote: Fei-Fei Li

NIPS Time Series Workshop 2017
Organizers include: Vitaly Kuznetsov
Authors include: Brendan Jou

OPT 2017: Optimization for Machine Learning
Organizers include: Sashank Reddi

ML Systems Workshop
Invited Speakers include: Rajat Monga, Alexander Mordvintsev, Chris Olah, Jeff Dean
Authors include: Alex Beutel, Tim Kraska, Ed H. Chi, D. Scully, Michael Terry

Aligned Artificial Intelligence
Invited Speakers include: Ian Goodfellow

Bayesian Deep Learning
Organizers include: Kevin Murphy
Invited speakers include: Nal Kalchbrenner, Matthew D. Hoffman

BigNeuro 2017
Invited speakers include: Viren Jain

Cognitively Informed Artificial Intelligence: Insights From Natural Intelligence
Authors include: Jiazhong Nie, Ni Lao

Deep Learning At Supercomputer Scale
Organizers include: Erich Elsen, Zak Stone, Brennan Saeta, Danijar Haffner

Deep Learning: Bridging Theory and Practice
Invited Speakers include: Ian Goodfellow

Interpreting, Explaining and Visualizing Deep Learning
Invited Speakers include: Been Kim, Honglak Lee
Authors include: Pieter Kinderman, Sara Hooker, Dumitru Erhan, Been Kim

Learning Disentangled Features: from Perception to Control
Organizers include: Honglak Lee
Authors include: Jasmine Hsu, Arkanath Pathak, Abhinav Gupta, James Davidson, Honglak Lee

Learning with Limited Labeled Data: Weak Supervision and Beyond
Invited Speakers include: Ian Goodfellow

Machine Learning on the Phone and other Consumer Devices
Invited Speakers include: Rajat Monga
Organizers include: Hrishikesh Aradhye
Authors include: Suyog Gupta, Sujith Ravi

Optimal Transport and Machine Learning
Organizers include: Olivier Bousquet

The future of gradient-based machine learning software & techniques
Organizers include: Alex Wiltschko, Bart van Merriënboer

Workshop on Meta-Learning
Organizers include: Hugo Larochelle
Panelists include: Samy Bengio
Authors include: Aliaksei Severyn, Sascha Rothe

Symposiums
Deep Reinforcement Learning Symposium
Authors include: Benjamin Eysenbach, Shane Gu, Julian Ibarz, Sergey Levine

Interpretable Machine Learning
Authors include: Minmin Chen

Metalearning
Organizers include: Quoc V Le

Competitions
Adversarial Attacks and Defences
Organizers include: Alexey Kurakin, Ian Goodfellow, Samy Bengio

Competition IV: Classifying Clinically Actionable Genetic Mutations
Organizers include: Wendy Kan

Tutorial
Fairness in Machine Learning
Solon Barocas, Moritz Hardt