Machine Learning Group

Established: August 3, 2016

Our current research focus is on deep/reinforcement learning, distributed machine learning, and graph learning. Other research projects from our group include learning to rank, computational advertising, and cloud pricing.

The Machine Learning Group at Microsoft Research Asia pushes the frontier of machine learning from theoretic, algorithmic, and practical aspects. Our current research focus is on deep/reinforcement learning, distributed machine learning, and graph learning. Other research projects from our group include learning to rank, computational advertising, and cloud pricing. We have published many highly-cited papers on top conferences and journals, helped our partner product groups apply machine learning to large and complex tasks, and open-sourced Microsoft Distributed Machine Learning Toolkit (DMTK) and Microsoft Graph Engine.

 


微软亚洲研究院机器学习组在理论、算法、应用等不同层面推动机器学习领域的学术前沿。我们目前的研究重点为深度学习/增强学习、分布式机器学习和图学习。我们的研究课题还包括排序学习、计算广告和云定价。在过去的十几年间,我们在顶级国际会议和期刊上发表了大量高质量论文,帮助微软的产品部门解决了很多复杂问题,并向开源社区贡献了微软分布式机器学习工具包(DMTK)和微软图引擎,并受到广泛关注。

Selected Publications

[Book]

  • Tie-Yan Liu. Learning to Rank for Information Retrieval, Springer, 2011.

[Journal Papers]

  • Shuaiqiang Wang, Shanshan Huang, Tie-Yan Liu, Jun Ma, Zhumin Chen, Jari Veijalainen, Ranking-oriented Collaborative Filtering: A Listwise Approach, ACM Transactions on Information Systems, 2016
  • Xujin Chen, Xiaodong Hu, Tie-Yan Liu, Weidong Ma, Tao Qin, Pingzhong Tang, Changjun Wang, and Bo Zheng, Efficient Mechanism Design for Online Scheduling, Journal of Artificial Intelligence Research, 2016.
  • Chang Xu, Gang Wang, Xiaoguang Liu, Tie-Yan Liu, Health Status Assessment and Failure Prediction for Hard Drives with Recurrent Neural Networks, IEEE Transactions on Computers, 2016.
  • Wei Chen, Tie-Yan Liu, and Xinxin Yang, Reinforcement Learning Behaviors in Sponsored Search, Applied Stochastic Models in Business and Industry, 2016.
  • Qing Cui, Bin Gao, Jiang Bian, Hanjun Dai, and Tie-Yan Liu, KNET: A General Framework for Learning Word Embedding using Morphological Knowledge, ACM Transactions on Information Systems, 2015.
  • Wei Wei, Bin Gao, Tie-Yan Liu, Taifeng Wang, Guohui Li, and Hang Li, A Ranking Approach on Large-scale Graph with Multi-dimensional Heterogeneous Information, IEEE Transactions on Cybernetics, 2015.
  • Tao Qin, Wei Chen, and Tie-Yan Liu, Sponsored Search Auctions: Recent Advances and Future Directions, ACM Transactions on Intelligent Systems and Technology, 2014.
  • Ying Zhang, Weinan Zhang, Bin Gao, Xiaojie Yuan, and Tie-Yan Liu, Bid Keyword Suggestion in Sponsored Search based on Competitiveness and Relevance, Information Processing and Management, 2014.
  • Di He, Wei Chen, Liwei Wang, and Tie-Yan Liu, Online Learning for Auction Mechanism in Bandit Setting, Decision Support Systems, 2013
  • Bin Gao, Tie-Yan Liu, Yuting Liu, Taifeng Wang, Zhiming Ma, and Hang Li, Page Importance Computation based on Markov Processes, Information Retrieval, 2011.
  • Olivier Chapelle, Yi Chang, and Tie-Yan Liu, Future research directions on learning to rank, Proceeding track, Journal of Machine Learning Research, 2011.
  • Xiubo Geng, Tie-Yan Liu, Tao Qin, Xueqi Cheng, Hang Li, Selecting Optimal Training Data for Learning to Rank, Information Processing and Management, 2011.
  • Yin He and Tie-Yan Liu, Tendency Correlation Analysis for Direct Optimization of Evaluation Measures in Information Retrieval, Information Retrieval, 2010.
  • Tie-Yan Liu, Thorsten Joachims, Hang Li, and Chengxiang Zhai, Introduction to special issue on learning to rank for information retrieval, Information Retrieval, 2010.
  • Tie-Yan Liu. Learning to Rank for Information Retrieval, Foundations and Trends in Information Retrieval, 2009.
  • Yuting Liu, Tie-Yan Liu, Zhiming Ma, and Hang Li. A Framework to Compute Page Importance based on User Behaviors, Information Retrieval, 2009.
  • Tao Qin, Tie-Yan Liu, and Hang Li, A General Approximation Framework for Direct Optimization of Information Retrieval Measures, Information Retrieval, 2009.
  • Tao Qin, Tie-Yan Liu, Jun Xu, and Hang Li, LETOR: A Benchmark Collection for Research on Learning to Rank for Information Retrieval, Information Retrieval, 2009
  • Tao Qin, Xu-Dong Zhang, Ming-Feng Tsai, De-Sheng Wang, Tie-Yan Liu, and Hang Li. Query-level Loss Function for Information Retrieval. Information Processing and Management, 2007.
  • Tao Qin, Xu-Dong Zhang, Tie-Yan Liu, De-Sheng Wang, Hong-Jiang Zhang. An Active Feedback Framework for Image Retrieval, Pattern Recognition Letters, 2007.
  • Ying Bao, Guang Feng, Tie-Yan Liu, Zhiming Ma and Ying Wang. Ranking Websites: A Probabilistic View, Internet Mathematics, 2007.
  • Tao Qin, Tie-Yan Liu, Xu-Dong Zhang, Guang Geng, De-Sheng Wang, and Wei-Ying Ma. Topic Distillation Via Subsite Retrieval, Information Processing and Management, 2006.
  • Bin Gao, Tie-Yan Liu, Xin Zheng, Qiansheng Cheng, and Wei-Ying Ma. Hierarchical Taxonomy Preparation for Text Categorization Using Consistent Bipartite Spectral Graph Co-partitioning, IEEE Transactions on Knowledge and Data Engineering, 2005.
  • Tie-Yan Liu, Yiming Yang, Hao Wan, Hua-Jun Zeng, Zheng Chen, and Wei-Ying Ma. Support Vector Machines Classification with Very Large Scale Taxonomy, SIGKDD Explorations, 2005.
  • Tie-Yan Liu, Kwoktung Lo, Xu-Dong Zhang, and Jian Feng. A New Cut Detection Algorithm with Constant False-Alarm Ratio for Video Segmentation, Journal of Visual Communications and Image Representation, 2004. 
  • Tie-Yan Liu, Xu-Dong Zhang, Jian Feng, and Kwoktung Lo. Shot Reconstruction Degree: a Novel Criterion for Key Frame Selection, Pattern Recognition Letters, 2004.
  • Tie-Yan Liu, Kwoktung Lo, Xu-Dong Zhang, and Jian Feng. Frame Interpolation Scheme Using Inertia Motion Prediction. Signal Processing: Image Communication, 2003.
  • Tie-Yan Liu, Kwoktung Lo, Xu-Dong Zhang, and Jian Feng. Inertia-based Cut Detection and Its Integration with Video Coder. IEE Proceedings on Vision, Image and Signal Processing, 2003.

[Conference Papers]

2016:

  • Xiang Li, Tao Qin, and Tie-Yan Liu, 2-Component Recurrent Neural Networks, NIPS 2016
  • Di He, Yingce Xia, Tao Qin, Tie-Yan Liu, and Wei-Ying Ma, Machine Translation Through Learning From a Communication Game, NIPS 2016
  • Qi Meng, Guolin Ke, Qiwei Ye, Taifeng Wang, Wei Chen, and Tie-Yan Liu, PV-Tree: A Communication-Efficient Parallel Algorithm for Decision Tree, NIPS 2016
  • Huazheng Wang, Fei Tian, Bin Gao, Chenjieren Zhu, Jiang Bian, Tie-Yan Liu, Solving Verbal Comprehension Questions in IQ Test by Knowledge-Powered Word Embedding, EMNLP 2016.
  • Yiren Wang, Fei Tian, Recurrent Residual Learning for Sequence Classification,  EMNLP 2016, short paper.
  • Yingce Xia, Tao Qin, Weidong Ma, Nenghai Yu and Tie-Yan Liu, Budgeted Multi-armed Bandits with Multiple Plays, IJCAI 2016.
  • Qi Meng, Wei Chen, Jingcheng Yu, Taifeng Wang and Tie-Yan Liu, Asynchronous Accelerated Stochastic Gradient Descent, IJCAI 2016.
  • Hongbin Ma, Bin Shao, Yanghua Xiao, Liang Jeff Chen, Haixun Wang. G-SQL: Fast Query Processing via Graph Exploration. PVLDB 2016.
  • Yingce Xia, Tao Qin, Tie-Yan Liu, Best Action Selection in a Stochastic Environment, AAMAS 2016.
  • Tie-Yan Liu, Weidong Ma, Pingzhong Tang, Tao Qin, Guang Yang, Bo Zheng, Online Non-Preemptive Story Scheduling in Web Advertising, AAMAS 2016
  • Jiang Rong, Tao Qin, Bo An, Tie-Yan Liu, Optimal Sample Size for Adword Auctions, AAMAS 2016.
  • Bo Zheng, Li Xiao, Guang Yang, Tao Qin, Online Posted-Price Mechanism with a Finite Time Horizon, AAMAS 2016, short paper.
  • Shizhao Sun, Wei Chen, Liwei Wang, and Tie-Yan Liu, On the Depth of Deep Neural Networks: A Theoretical View, AAAI 2016.

2015:

  • Yingce Xia, Haifang Li, Tao Qin, Nenghai Yu, and Tie-Yan Liu, Thompson Sampling for Budgeted Multi-armed Bandits, IJCAI 2015.
  • Bolei Xu, Tao Qin, Guoping Qiu, and Tie-Yan Liu, Competitive Pricing for Cloud Computing in an Evolutionary Market, IJCAI 2015.
  • Changjun Wang, Weidong Ma, Tao Qin, Xujin Chen, Xiaodong Hu, and Tie-Yan Liu, Selling Reserved Instances in Cloud Computing, IJCAI 2015.
  • Long Tran-Thanh, Yingce Xia, Tao Qin, Nick Jenning, Efficient Algorithms with Performance Guarantees for the Stochastic Multiple-Choice Knapsack Problem, IJCAI 2015.
  • Yitan Li, Linli Xu, Fei Tian, Liang Jiang, Xiaowei Zhong and Enhong Chen, Word Embedding Revisited: A New Representation Learning and Explicit Matrix Factorization Perspective, IJCAI 2015.
  • Shanshan Huang, Shuaiqiang Wang, Tie-Yan Liu, Jun Ma, Zhumin Chen, and Jari Veijalainen, Listwise Collaborative Filtering, SIGIR 2015.
  • Binyi Chen, Tao Qin, and Tie-Yan Liu, Mechanism Design for Daily Deals, AAMAS 2015.
  • Jiang Rong, Tao Qin, and Bo An. Quantal Response Equilibrium for Sponsored Search Auctions, AAMAS 2015, short paper.
  • Jinhui Yuan, Fei Gao, Qirong Ho, Wei Dai, Jinliang Wei, Xun Zheng, Eric Xing, Tie-Yan Liu, and Wei-Ying Ma, LightLDA: Big Topic Models on Modest Computer Cluster, WWW 2015.
  • Tie-Yan Liu, Wei Chen, and Tao Qin, Mechanism Learning with Mechanism Induced Data, Senior Member Track, AAAI 2015.
  • Haifang Li, Wei Chen, Fei Tian, Tao Qin, and Tie-Yan Liu, Generalization Analysis for Game-theoretic Machine Learning, AAAI 2015.
  • Qizhen Zhang, Haoran Wang, Yang Chen, Tao Qin, Ying Yan, Thomas Moscibroda, A Shapley Value Approach for Cost Allocation in the Cloud, SOCC 2015, poster.
  • Liang He, Bin Shao, Yatao Li, Enhong Chen. Distributed Real-Time Knowledge Graph Serving. BigComp 2015. Invited Paper.

2014:

  • Chang Xu, Yalong Bai, Jiang Bian, Bin Gao, and Tie-Yan Liu, A General Approach to Incorporate Knowledge into Word Representation, CIKM 2014.
  • Fei Tian, Jiang Bian, Bin Gao, Hanjun Dai, Rui Zhang, and Tie-Yan Liu, A Scalable Probabilistic Model for Learning Multi-Prototype Word Embedding, COLING 2014.
  • Siyu Qiu, Qing Cui, Jiang Bian, Bin Gao, and Tie-Yan Liu, Co-learning of Word Representations and Morpheme Representations, COLING 2014.
  • Bin Gao, Jiang Bian, and Tie-Yan Liu, Knowledge Powered Deep Learning for Word Embedding, ECML/PKDD 2014.
  • Junpei Komiyama and Tao Qin, Time-Decaying Bandits for Non-stationary Systems, WINE 2014.
  • Wei Chen, Di He, Tie-Yan Liu, Tao Qin, Yixin Tao, Liwei Wang, Generalized Second Price Auction with Probabilistic Broad Match, EC 2014.
  • Jiang Rong, Tao Qin, and Bo An. Quantal Response Equilibrium for Sponsored Search Auctions: Computation and Inference, Ad Auctions 2014, in conjunction with EC 2014.
  • Yingce Xia, Tao Qin and Tie-Yan Liu, Incentivizing High-quality Content from Heterogeneous Users: On the Existence of Nash Equilibrium, AAAI 2014.
  • Fei Tian, Haifang Li, Wei Chen, Tao Qin and Tie-Yan Liu, Agent Behavior Prediction and Its Generalization Analysis, AAAI 2014.
  • Fei Tian, Bin Gao and Tie-Yan Liu, Learning Deep Representations for Graph Clustering, AAAI 2014.
  • Yuyu Zhang, Hanjun Dai, Chang Xu, Jun Feng, Taifeng Wang, Jiang Bian, Bin Wang and Tie-Yan Liu, Sequential Click Prediction for Sponsored Search with Recurrent Neural Networks, AAAI 2014.
  • Tie-Yan Liu, Weidong Ma, Tao Qin, and Tao Wu, Generalized Second Price Auctions with Value Externalities, AAMAS 2014.
  • Jiang Bian, Taifeng Wang, and Tie-Yan Liu, Sampling Dilemma: Towards Effective Data Sampling for Click Prediction in Sponsored Search, WSDM 2014.
  • Lu Wang, Yanghua Xiao, Bin Shao, Haixun Wang, How to Partition a Billion-Node Graph ICDE 2014.
  • Huanhuan Xia, Tun Lu, Bin Shao, Guo Li, Xianghua Ding, Ning Gu, A partial Replication Approach for Anywhere Anytime Mobile Commenting   CSCW 2014.
  • Zichao Qi, Yanghua Xiao, Bin Shao, Haixun Wang. Toward a Distance Oracle for Billion-Node Graphs. PVLDB 2014.

2013:

  • Yining Wang, Liwei Wang, Yuanzhi Li, Di He, Wei Chen, and Tie-Yan Liu, A Theoretical Analysis of NDCG Type Ranking Measures, COLT 2013.
  • Weihao Kong, Jian Li, Tie-Yan Liu and Tao Qin, Optimal Allocation for Chunked-Reward Advertising, WINE 2013.
  • Min Xu, Tao Qin, and Tie-Yan Liu, Estimation Bias in Multi-Armed Bandit Algorithms for Search Advertising, NIPS 2013.
  • Taifeng Wang, Jiang Bian, Shusen Liu, Yuyu Zhang, and Tie-Yan Liu, Psychological Advertising: Exploring Consumer Psychology for Click Prediction in Sponsored Search, KDD 2013.
  • Bin Shao, Haixun Wang, Yatao Li, Trinity: A Distributed Graph Engine on a Memory Cloud, SIGMOD 2013.  
  • Kai Zeng, Jiacheng Yang, Haixun Wang, Bin Shao, Zhongyuan Wang, A Distributed Graph Engine for Web Scale RDF Data, PVLDB 2013.
  • Di He, Wei Chen, Liwei Wang, and Tie-Yan Liu, A Game-theoretic Machine Learning Approach for Revenue Maximization in Sponsored Search, IJCAI 2013.
  • Wenkui Ding, Tao Qin, and Tie-Yan Liu, Multi-Armed Bandit with Budget Constraint and Variable Costs, AAAI 2013.
  • Haifeng Xu, Diyi Yang, Bin Gao and Tie-Yan Liu, Predicting Advertiser Bidding Behaviors in Sponsored Search by Rationality Modeling, WWW 2013.

2012:

  • Lei Yao, Wei Chen and Tie-Yan Liu, Convergence Analysis for Weighted Joint Strategy Fictitious Play in Generalized Second Price Auction, WINE 2012.
  • Weinan Zhang, Ying Zhang, Bin Gao, Yong Yu, Xiaojie Yuan, and Tie-Yan Liu, Joint optimization of bid and budget allocation in sponsored search, KDD 2012.
  • Chenyan Xiong, Taifeng Wang, Wenkui Ding, Yidong Shen, Tie-Yan Liu. Relational Click Prediction for Sponsored Search, WSDM 2012.
  • Yanyan Lan, Jiafeng Guo, Xueqi Cheng, Tie-Yan Liu, Statistical Consistency of Ranking Methods in A Rank-Differentiable Probability Space. NIPS 2012.
  • Zhao Sun, Hongzhi Wang, Haixun Wang, Bin Shao, and Jianzhong Li. Efficient Subgraph Matching on Billion Node Graphs. PVLDB 2012

2011:

  • Bin Gao, Tie-Yan Liu, Taifeng Wang, Wei Wei, and Hang Li, Semi-supervised graph ranking with rich meta data, KDD 2011
  • Zhicong Cheng, Bin Gao, Congkai Sun, Yanbing Jiang, and Tie-Yan Liu. Let Web Spammers Expose Themselves, WSDM 2011.

2010:

  • Zhicong Cheng, Bin Gao, and Tie-Yan Liu, Actively Predicting Diverse Search Intent from User Browsing Behaviors, WWW 2010.
  • Tao Qin, Xiubo Geng, and Tie-Yan Liu, A New Probabilistic Model for Rank Aggregation, NIPS 2010.
  • Wei Chen, Tie-Yan Liu, Zhiming Ma, Two-Layer Generalization Analysis for Ranking Using Rademacher Average, NIPS 2010.
  • Jiang Bian, Tie-Yan Liu, Tao Qin, and Hongyuan Zha, Query-dependent Loss Function for Web Search. WSDM 2010.

Pre-2010:

  • Fen Xia, Tie-Yan Liu, Hang Li, Statistical Consistency of Top-k Ranking, NIPS 2009.
  • Wei Chen, Tie-Yan Liu, Yanyan Lan, Zhiming Ma, Hang Li, Ranking Measures and Loss Functions in Learning to Rank, NIPS 2009.
  • Yanyan Lan, Tie-Yan Liu, Zhiming Ma, and Hang Li. Generalization Analysis for Listwise Learning to Rank Algorithms, ICML 2009.
  • Tao Qin, Tie-Yan Liu, Xu-Dong Zhang, De-Sheng Wang, and Hang Li. Global Ranking Using Continuous Conditional Random Fields, NIPS 2008.
  • Fen Xia, Tie-Yan Liu, Jue Wang, Wensheng Zhang, and Hang Li. Listwise Approach to Learning to Rank: Theory and Algorithm, ICML 2008.
  • Yanyan Lan, Tie-Yan Liu, Tao Qin, Zhiming Ma, and Hang Li. Query-level Stability and Generalization in Learning to Rank, ICML 2008.
  • Tao Qin, Tie-Yan Liu, Xu-Dong Zhang, De-Sheng Wang, Wen-Ying Xiong, and Hang Li. Learning to Rank Relational Objects and Its Application to Web Search, WWW 2008.
  • Xiubo Geng, Tie-Yan Liu, Tao Qin, Andrew Arnold, Hang Li, and Heung-Yeung Shum. Query-dependent Ranking using K-Nearest Neighbor, SIGIR 2008.
  • Yuting Liu, Bin Gao, Tie-Yan Liu, Ying Zhang, Zhiming Ma, Shuyuan He, and Hang Li. BrowseRank: Letting Web Users Vote for Page Importance, SIGIR 2008.
  • Jun Xu, Tie-Yan Liu, Min Lu, Hang Li, and Wei-Ying Ma. Directly Optimizing IR Evaluation Measures in Learning to Rank, SIGIR 2008.
  • Zhe Cao, Tao Qin, Tie-Yan Liu, Ming-Feng Tsai, and Hang Li. Learning to Rank: From Pairwise Approach to Listwise Approach. ICML 2007.
  • Yuting Liu, Tie-Yan Liu, Tao Qin, Zhi-Ming Ma, and Hang Li. Supervised Rank Aggregation, WWW 2007.
  • Xiubo Geng, Tie-Yan Liu, Tao Qin, and Hang Li. Feature Selection for Ranking, SIGIR 2007.
  • Mingfeng Tsai, Tie-Yan Liu, Tao Qin, Hsin-Hsi Chen, and Wei-Ying Ma. FRank: A Ranking Method with Fidelity Loss, SIGIR 2007.
  • Tao Qin, Tie-Yan Liu, Wei Lai, Xu-Dong Zhang, De-Sheng Wang, and Hang Li. Ranking with Multiple Hyperplanes, SIGIR 2007.
  • Guang Feng, Tie-Yan Liu, Ying Wang, Ying Bao, Zhiming Ma, Xu-Dong Zhang, and Wei-Ying Ma. AggregateRank: Bringing Order to Websites, SIGIR 2006.
  • Yunbo Cao, Jun Xu, Tie-Yan Liu, Hang Li, Yalou Huang, and Hsiao-Wuen Hon. Adapting Ranking SVM to Document Retrieval, SIGIR 2006.
  • Qiankun Zhao, Chuhong Hoi, Tie-Yan Liu, Sourav S. Bhowmick, Michael R. Lyu, and Wei-Ying Ma. Time-Dependent Semantic Similarity Measure of Queries Using Historical Click-Through Data, WWW 2006.
  • Qiankun Zhao, Tie-Yan Liu, Sourav S. Bhowmick, and Wei-Ying Ma. Event Detection from Evolution of Click-through Data, KDD 2006.
  • Tao Qin, Tie-Yan Liu, Xu-Dong Zhang, Zheng Chen, and Wei-Ying Ma. A Study on Relevance Propagation for Web Search, SIGIR 2005.
  • Bin Gao, Tie-Yan Liu, Xin Zheng, Qian-Sheng Cheng, and Wei-Ying Ma. Consistent Bipartite Graph Co-Partitioning for Star-Structured High-Order Heterogeneous Data Co-Clustering, KDD 2005.
  • Bin Gao, Tie-Yan Liu, Tao Qin, Xin Zheng, Qian-Sheng Cheng, and Wei-Ying Ma. Web Image Clustering by Consistent Utilization of Visual Features and Surrounding Texts, ACM Multimedia 2005.
  • Tie-Yan Liu, Tao Qin and Hong-Jiang Zhang. Time-constraint Boost for TV Commercials Detection. IEEE ICIP 2004.
  • Bin Gao, Tie-Yan Liu, Qian-Sheng Cheng, and Wei-Ying Ma. A Linear Approximation Based Method for Noise-Robust and Illumination-invariant Image Change Detection. PCM 2004.

People

Projects

Reinforcement learning for Internet applications

Established: August 1, 2016

Reinforcement learning (RL) has achieved great success in video and board games, in which RL can make improvement through (almost) infinite self-play. However, different from games, in real-world applications,  self-play is either impossible or costly. In this project, our goal is to make RL feasible and successful in real-world applications. Our targeted applications include machine translation, chat bot, and search advertising.

Reinforcement learning for machine learning

Established: August 1, 2016

Reinforcement learning (RL) has achieved great success in video and board games. In this project, we aim at boosting machine learning algorithms and systems by leveraging reinforcement learning techniques. We focus the following aspects. First, RL for data selection and pre-processing, in which we use RL techniques to select right data at right time and process the data in a right way for model training. Second, RL for hyper parameter optimization. Setting appropriate hyper parameters…

Mechanism design for complex systems

Established: August 1, 2016

We are living in and interacting with many complex systems nowadays, including the advertising system, cloud system, large scale machine learning system, and multitenancy big data system. There are multiple kinds of players with different objectives and there could be millions of players in total. Designing good mechanisms is important for those systems to achieve efficiency, revenue, and stability. We combine machine learning, user modeling, and game theory techniques together to design mechanisms for those…

Graph Engine

Established: August 1, 2016

Graph Engine, previously known as Trinity, is a distributed, in-memory, large graph processing engine:  http://www.graphengine.io/

Symbolic Learning

Established: August 1, 2016

The goal of this project is to build a symbolic knowledge service stack, providing the capability of knowledge storage and serving as well as knowledge learning and reasoning. This project tries to represent knowledge using a unified formal symbolic graph and conduct symbolic inference based on graph matching and graph transformation.

LIKQ

Established: August 1, 2016

LIKQ is a Language-Integrated Knowledge Query language. It allows users to query, search, and consume knowledge via graph traversal and lambda expressions in real time, making massive knowledge accessible at our fingertips.

LightLDA

Established: August 1, 2016

LightLDA is a distributed system for large scale topic modeling. It implements a distributed sampler that enables very large data sizes and models. LightLDA improves sampling throughput and convergence speed via a fast O(1) metropolis-Hastings algorithm, and allows small cluster to tackle very large data and model sizes through model scheduling and data parallelism architecture. LightLDA is implemented with C++ for performance consideration. We have sucessfully trained big topic models (with trillions of parameters) on…

GPU based Linear Programming

Established: August 1, 2016

This project is to develop a highly parallel simplex method for solving linear programming problems on top of GPU architecture

PV-Tree – parallel GBDT

Established: August 1, 2016

GBDT is a widely used machine learning tool in the industry practice. This project solves the parallelization problem of this algorithm. We proposed a local/global voting based method call PV-Tree, which dramatically reduces the communication cost of parallel GBDT training and leads to great efficiency of our parallel GBDT algorithm. The work in this project includes the following aspects: 1) a new parallel version of FastRank, which is the most widely used ranking/classification tool used…

Multiverso – Parameter server platform for distribute machine learning

Established: August 1, 2016

Parameter server based distributed machine learning has been widely adopted in many distributed machine learning platform. This project is targeting to build the best parameter server framework for distributed machine learning. Research in this project includes: 1) Design flexible and efficient parameter server interfaces which can empower distributed training of existing machine learning algorithms with just a few lines of additional code. 2) Advance the distribute optimization algorithms to produce better parallel training result, e.g.…

Distributed DNN platform

Established: August 1, 2016

DNN is a really important and practical machine learning capability,  this project focuses on finding solution for distributing the DNN training on a cluster of machines based on our parameter server framework. The research in this project include: 1) support very efficient distributed training of DNN in MS CNTK project by introducing important features like asynchronous training, efficient sparse model training, GPU based optimization, rich NN related algorithms, model parallelism for training super big model.…

Model scheduling for super large scale model training

Established: August 1, 2016

Machine learning in this big data era is facing two challenges, one is big data, the other is big model (one machine cannot hold the entire model in runtime). In distributed machine learning, people has proposed data parallelism to solve the big data problem by partitioning the data to different machines while training many replicas of the same model simultaneously. As for big model problem, model parallelism is proposed to solve it by partition the…

Click prediction in sponsored search

Established: August 1, 2016

Click prediction is one of the major technical parts in sponsored search which dramatically affect the overall revenue of search engine. In this project we have proposed quite a few new method and aspects to improve the click prediction model in sponsored search. And many of them have been used in Bing search.

DMTK

Established: August 1, 2016

DMTK is an open source project running in our group to share our recent research output on distributed machine learning with open source community.  It contains our Multiverso parameter server platform and many of our distributed algorithms like distributed DNN, LR, Word Embedding, LDA, GBDT. Machine learning practitioners can easily leverage our powerful distributed machine learning algorithm to deal with their machine learning problems. What’s more, with this parameter server infrastructure, machine learning researchers can…

Machine Learning Theory

Established: August 1, 2016

Theory plays an important role in improving existing algorithms and motivating new algorithms. In this project, we aim to give mathematical formulations, derive theoretical results, and design new algorithms to variants of machine learning problems. In the past several years, we have investigated or are investigating the following topics: Distributed machine learning algorithms Deep neural networks Game-theoretic machine learning Learning to rank