Posted in: AI

IBM Research showcases AI advances @ NIPS 2017

At the 2017 NIPS conference in Long Beach, CA, IBM will showcase new advances from its AI research team via technical papers as well as results from the company’s ongoing collaboration with academic institutions through the MIT IBM Watson AI Lab and the AI Horizons Network.AI

IBM and MIT scientists will unveil and publish a paper detailing the Moments in Time Dataset, a large-scale dataset of one million three-second annotated video clips that can be used to accelerate the development of technologies and models that will enable automatic video understanding and can be applied in diverse domains such as assisting the visually impaired, elderly care, automotive, media & entertainment, and more.

At the conference, IBM scientists will disclose technical progress in machine learning, computer vision, neural network performance and other core AI research areas, such as significant scientific results in reducing bias contained in data on which AI algorithms are trained and delivering more relevant product purchase recommendations to online consumers.

Among the featured IBM AI research papers that will be detailed at NIPS is pioneering work from a team of IBM researchers who invented an AI algorithm that can enable developers to train large-scale machine learning models 10x faster than existing methods. The researchers demonstrated the speed and efficiency of their new algorithm by training 40,000 photos of cats and dogs in less than one minute. Another paper outlines an AI system created by IBM scientists that used machine learning and the state of the art in language translation to learn organic chemistry and design completely new chemical reactions not yet discovered by human chemists.

IBM’s robust NIPS presence illustrates how the company is applying its investment in foundational AI research to produce innovations in AI engineering, science and technology to solve real business challenges, while also exploring and understanding how AI can be effectively applied to solve societal problems.

Below is a complete look at the papers and workshops IBM Research will present at NIPS. If you’re in Long Beach, come visit us at booth 215.

Papers

Authors: Vernon Austel (IBM Research), Sanjeeb Dash (IBM Research), Oktay Gunluk (IBM Research), Lior Horesh (IBM Research), Leo Liberti (Ecole Polytechnique), Giacomo Nannicini (IBM Research), Baruch Schieber (IBM Research)  

Learn more at NIPS:

Thursday December 7, 3:10-3:40 PM @ Hall C

Interpretable and Globally Optimal Prediction for Textual Grounding using Image Concepts 

Authors: Raymond Yeh and Alexander Schwing (University of Illinois at Urbana–Champaign); Jinjun Xiong (IBM Research); Wen-Mei Hwu and Minh Do (University of Illinois)

Learn more at NIPS:

Tuesday December 5, 4:35-4:50 PM @ Hall A (oral presentation)

Tuesday December 5, 6:30-10:30 PM @ Pacific Ballroom #82 (poster)

 

Can Decentralized Algorithms Outperform Centralized Algorithms? A Case Study for Decentralized Parallel Stochastic Gradient Descent

Authors: Xiangru Lian (University of Rochester); Ce Zhang (ETH Zurich), Huan Zhang and Cho-Jui Hsieh (UC Davis); Wei Zhang (IBM Research); Ji Liu (University of Rochester)

Learn more at NIPS:

Wednesday December 6, 3:05-3:20 PM @ Hall C (oral presentation)

Wednesday December 6, 6:30-10:30 PM @ Pacific Ballroom #167 (poster)

 

Communication-Efficient Distributed Learning of Discrete Distributions

Authors: Ilias Diakonikolas, Elena Grigorescu and Abhiram Natarajan (Purdue University); Jerry Li and Ludwig Schmidt (MIT), Krzysztof Onak (IBM Research)

Learn more at NIPS:

Tuesday December 5, 4:50-5:05 PM @ Hall C (oral presentation)

Tuesday December 5, 6:30-10:30 PM @ Pacific Ballroom #61 (poster)

 

Net-Trim: Convex Pruning of Deep Neural Networks with Performance Guarantee

Authors: Alireza Aghasi and Nam Nguyen (IBM Research); Justin Romberg (Georgia Institute of Technology)

Learn more at NIPS:

Tuesday December 5, 3:25-3:30 PM @ Hall C (spotlight)

Tuesday December 5, 6:30-10:30 PM @ Pacific Ballroom #215 (poster)

 

Efficient Use of Limited-Memory Resources to Accelerate Linear Learning

Authors: Celestine Dünner and Thomas Parnell (IBM Research); Martin Jaggi (EPFL)

Learn more at NIPS:

Monday December 4, 6:30-10:30 PM @ Pacific Ballroom #175 (poster)

 

Optimized Pre-Processing for Discrimination Prevention

Authors: Flavio Calmon (Harvard University); Dennis Wei, Karthikeyan Ramamurthy, Bhanukiran Vinzamuri and Kush R Varshney (IBM Research)

Learn more at NIPS:

Wednesday December 6, 6:30-10:30 PM @ Pacific Ballroom #76 (poster)

 

Dilated Recurrent Neural Networks

Authors: Shiyu Chang, Yang Zhang, Xiaoxiao Guo, Wei Tan, Xiaodong Cui and Michael Witbrock (IBM Research); Wei Han (University of Illinois at Urbana-Champaign); Mo Yu (Johns Hopkins University); Mark A Hasegawa-Johnson (University of Illinois); Thomas Huang (UIUC)

Learn more at NIPS:

Monday December 4, 6:30-10:30 PM @ Pacific Ballroom #104

 

Mixture-Rank Matrix Approximation for Collaborative Filtering

Authors: Dongsheng Li and Stephen Chu (IBM Research); Chao Chen (Tongji University); Wei Liu (Tencent Technology (Shenzhen) Company Limited); Tun Lu and Ning Gu (Fudan University)

Learn more at NIPS:

Monday December 4, 6:30-10:30 PM @ Pacific Ballroom #47

 

Improved Dynamic Regret for Non-degeneracy Functions

Authors: Lijun Zhang, Rong Jin and Zhi-Hua Zhou (Nanjing University); Tianbao Yang (The University of Iowa); Jinfeng Yi (IBM Research)

Learn more at NIPS:

Tuesday December 5, 6:30-10:30 PM @ Pacific Ballroom #62

 

MMD GAN: Towards Deeper Understanding of Moment Matching Network

Authors: Chun-Liang Li, Wei-Cheng Chang and Barnabas Poczos (Carnegie Mellon University); Yu Cheng (IBM Research); Yiming Yang (CMU)

Learn more at NIPS:

Wednesday December 6, 6:30-10:30 PM @ Pacific Ballroom #107

 

Scalable Demand-Aware Recommendation

Authors: Jinfeng Yi and Kush R Varshney (IBM Research); Cho-Jui Hsieh and Yao Li (UC Davis); Lijun Zhang (Nanjing University)

Learn more at NIPS:

Wednesday December 6, 6:30-10:30 PM @ Pacific Ballroom #88

 

Fisher GAN

Authors: Youssef Mroueh and Tom Sercu (IBM Research)

Learn more at NIPS:

Wednesday December 6, 6:30-10:30 PM @ Pacific Ballroom #104

 

Model-Powered Conditional Independence Test

Authors: Rajat Sen (University of Texas at Austin); Ananda Theertha Suresh (Google); Karthikeyan Shanmugam (IBM Research); Alexandros Dimakis and Sanjay Shakkottai (The University of Texas at Austin)

Learn more at NIPS:

Monday December 4, 6:30-10:30 PM @ Pacific Ballroom #190

 

Wasserstein Learning of Deep Generative Point Process Models

Authors: Shuai Xiao (Georgia Institute of Technology); Mehrdad Farajtabar, Le Song and Hongyuan Zha (Georgia Tech); Xiaojing Ye (Georgia State University); Junchi Yan (IBM Research)

Learn more at NIPS:

Monday December 4, 6:30-10:30 PM @ Pacific Ballroom #106

 

Improved Semi-supervised Learning with GANs using Manifold Invariances

Authors: Abhishek Kumar and Prasanna Sattigeri (IBM Research); Tom Fletcher (University of Utah)

Learn more at NIPS:

Wednesday December 6, 6:30-10:30 PM @ Pacific Ballroom #110

 

Learning Causal Graphs with Latent Variables

Murat Kocaoglu (University of Texas at Austin); Karthikeyan Shanmugam (IBM Research); Elias Bareinboim (Purdue)

Learn more at NIPS:

Wednesday December 6, 6:30-10:30 PM @ Pacific Ballroom #184

 

“Found in translation”: Predicting Outcomes of Complex Organic Chemistry Reactions Using Neural Sequence-to-sequence Models

Philippe Schwaller, Théophile Gaudin, Dávid Lányi, Costas Bekas, Teodoro Laino

Learn more at NIPS:

Friday December 8, 8:00 – 5:55 PM @ Room 102-C

 

Workshops, Friday December 8

Time Series 
8:00 AM-6:30 PM @ Grand Ballroom A
Dynamic Boltzmann machines for second order moments and generalized Gaussian distributions
Takayuki Osogamio

Fluid simulation with dynamic Boltzmann machine in batch manner
Kun Zhao, Takayuki Osogami

Learning with Limited Labeled Data: Weak Supervision and Beyond Workshop
8:00 AM-6:30 PM @ Grand Ballroom B
Learning Loss Functions for Semi-supervised Learning via Discriminative Adversarial Networks
Cicero dos Santos, Kahini Wadhawan, Bowen Zhou

Machine Learning for the Developing World
8:00 AM-6:30 PM @ S7
Neurology-as-a-Service for the Developing World
Tejas Dharamsi, Payel Das, Tejaswini Pedapati, Gregory Bramble, Vinod Muthusamy, Horst Samulowitz, Kush R. Varshney

Automated Knowledge Base Creation
8:00 AM-6:30 PM @ 102 C
Extending Knowledge Base with Images
Vincent Lonij, Ambrish Rawat, Maria-Irina Nicolae

Harnessing Model Uncertainty for Adversarial Detection
Ambrish Rawat, Martin Wistuba, Maria-Irina Nicolae

Optimization for Machine Learning
8:30 AM-6:30 PM @ 510 ac
Convergence Analysis of Zeroth-Order Online Alternating Direction Method of Multipliers
Sijia Liu, Pin-Yu Chen, Jie Chen

Low-Rank Boolean Matrix Approximation by Integer Programming
Oktay Gunluk (IBM Research), Raphael Hauser and Reka Agnes Kovacs (University of Oxford)

Machine Learning and Computer Security
8:00 AM-6:30 PM @ Hyatt Shoreline
ZOO: Zeroth Order Optimization based Black-box Attacks to Deep Neural Networks without Training Substitute Models
Pin-Yu Chen

Machine Learning for Creativity and Design
8:00 AM-6:30 PM @ Hyatt Hotel Seaview Ballroom
AI for Fragrance Design
Richard Goodwin, Joana Maria, Payel Das, Raya Horesh, Richard Segal, Jing Fu, Christian Harris

Machine Learning for Health
8:00 AM-6:30 PM @ Room 116
Learning Neural Markers of Schizophrenia Disorder Using Recurrent Neural Networks
Irina Rish

Machine Learning for Molecules and Materials
8:00 AM-6:30 PM @ Room 102-C
“Found in translation”: Predicting Outcomes of Complex Organic Chemistry Reactions Using Neural Sequence-to-sequence Models [arXiv]
Philippe Schwaller, Théophile Gaudin, Dávid Lányi, Costas Bekas, Teodoro Laino

Workshop on ML Systems
8:00 AM-6:30 PM @ S1
Scalable Multi-Framework Multi-Tenant Lifecycle Management of Deep Learning Training Jobs
Scott Boag, Parijat Dube, Benjamin Herta, Waldemar Hummer, Vatche Ishakian, K. R. Jayaram, Michael Kalantar, Vinod Muthusamy, Priya Nagpurkar, Florian Rosenberg

Workshops, Saturday December 9

Bayesian Deep Learning
8:00 AM-6:30 PM @ Hall C
Model Selection in Bayesian Neural Networks via Horseshoe Priors
Soumya Ghosh

Learning Disentangled Representations: from Perception to Control
8:00 AM-6:30 PM @ 203
Variational Inference of Disentangled Latents from Unlabeled Observations
Abhishek Kumar, Prasanna Sattigeri, Avinash Balakrishnan

Improved Neural Text Attribute Transfer with Non-parallel Data
Igor Melnyk, Cicero dos Santos, Kahini Wadhawan, Inkit Padhi, Abhishek Kumar

Hierarchial Reinforcement Learning
8:00 AM-6:30 PM @ Grand Ballroom A
The Eigenoption Critic Framework
Miao Liu, Gerald Tesauro, Murray S. Campbell

Teaching Machines, Robots, and Humans
8:00 AM-6:30 PM @ Hyatt Hotel Seaview Ballroom
Generative Knowledge Distillation for General Purpose Function Compression
Matthew Reimer, Michele Franceschini, Djallel Bouneffouf, Tim Klinger

Deep Learning: Bridging Theory and Practice
8:00 AM-6:30 PM @ Room 112
Semi-Supervised Learning with IPM-based GANs: an Empirical Study
Tom Sercu, Youssef Mroueh

Workshop on Metalearning
8:00 AM-6:30 PM @ Hyatt Hotel Beacon Ballroom
Selection of Non-Linear Function for Multi-Task Learning
Tim Klinger, Matthew Reimer

Learning with Limited Labeled Data: Weak Supervision and Beyond
8:00 AM-6:30 PM @ Grand Ballroom B
Bootstrapping Chatbots for Novel Domains
Petr Babkin, Md Faisal Chowdhury, Alfio Gliozzo, Martin Hirzel, Avraham Shinnar

Few-Shot Learning with Meta Metric Learners
Yu Cheng, Mo Yu, Xiaoxiao Guo, Bowen Zhou

 

Comments