Conference
Yandex School of Data Analysis Conference Machine Learning: Prospects and Applications
5–8 October 2015, Berlin, Germany
The conference is organised by the Yandex School of Data Analysis and Yandex, sponsored by Yandex Data Factory. Royal Holloway, University of London is the academical partner of the conference.
This is the second conference on machine learning organised by the Yandex School of Data Analysis and this year it will explore the new frontiers and advances in machine learning – its theory, technology, and applications. With scientists and engineers coming from 14 countries, in addition to talks, discussions and poster sessions, this year’s conference will feature a business event aimed at facilitating cooperation between academics and practitioners. The first conference on machine learning organised by the Yandex School of Data Analysis, which took place in Moscow in 2013, focused on the latest developments in big data.
Over the past few years, deep learning has been gaining popularity among software developers for the extraordinary experimental results it has demonstrated. More importantly, using this approach enabled developers to create simple tools for solving practical problems across multiple industries. A number of speakers at this conference will talk about this approach in their work. At the same time, Vladimir Vapnik, one of the main developers of Vapnik–Chervonenkis theory, the father of statistical learning theory, and the co-inventor of the Support Vector Machine method, over the course of the past two years, has had tremendous theoretical progress in his breakthrough theory of machine learning, which he called Intelligent Learning. This novel approach will be presented in detail at the conference by the author and his colleagues. This conference offers a unique opportunity for everyone in attendance to witness first-hand how these two alternative approaches compare and contrast. During a specially organised discussion, we are hoping to open new advantages of both of these approaches and see the power of synergy stemming from their interrelations. Beside looking at the exciting prospects promised by deep learning and Vapnik’s Intelligent learning, the conference will explore traditional application areas for machine learning, such as computer vision or natural language processing, alongside some relatively recent ones, such as causality modeling. As significant progress has been made in both old and new areas of application for machine learning, some of the most important cases will be presented at this conference.
The main topics
- Deep Learning
- Intelligent Learning
- Discovery of Causality
- Abstract Convexity
- Sub-Linear Methods in Data Analysis
- Quantum Calculation in Machine Learning
- Text Analysis and Understanding
- Video, Image and Signal Analysis
- Application in Physics, Biology, Medicine and Finance
Program
October 5
Begin
| October 5 | Registration |
| 8:30 – 8:40 | Welcome & Keynote Address Prof. Elena Bunina Russia, Yandex, Yandex School of Data Analysis and Moscow State University Prof. Ilya Muchnik USA, Rutgers University (NJ), and Russia, Yandex School of Data Analysis |
Session I
Chair Sc.D. Anatoli Michalski, RussiaV. A. Trapeznikov Institute of Control Sciences of Russian Academy of Sciences, Moscow
Deep Learning | |
| 8:40 – 9:20 | Image Annotation – The Marriage of Computer Vision and NLP Using Deep Learning Prof. Lior Wolf Israel, Blavatnik School of Computer Science at Tel Aviv University |
| 9:20 – 10:00 | Deep Generative and Discriminative Models for Speech Recognition Dr. Li Deng USA, Microsoft Research |
| 10:00 – 10:15 | Break |
| 10:15 – 10:55 | Deep-er Kernels Prof. John Shawe-Taylor UK, Centre for Computation Statistics and Machine Learning at University College London |
| 10:55 – 11:35 | Deep Learning Applications in the Natural Sciences Deep Learning Applications in the Natural Sciences USA, Department of Computer Science, Center for Machine Learning and Intelligent Systems, and Institute for Genomics and Bioinformatics at University of California, Irvine |
| 11:35 – 11:50 | Break |
Sublinearity | |
| 11:50 – 12:10 | Sublinear-time Approximation Algorithms Prof. Artur Czumaj UK, Centre for Discrete Mathematics and its Applications (DIMAP), University of Warwick |
Generalizations | |
| 12:10 – 12:30 | Data Properties Estimation as Statistical Inverse Problem Sc.D. Anatoli Michalski Russia, V. A. Trapeznikov Institute of Control Sciences of Russian Academy of Sciences, Moscow |
| 12:30 – 12:50 | Feature Selection by Conditional Distribution Contrasting Dr. Varvara Tsurko Russia, V. A. Trapeznikov Institute of Control Sciences of Russian Academy of Sciences, Moscow |
| 12:50 – 13:40 | Lunch |
Session II
Chair Arkady Borkovsky, USAYandex
Causality | |
| 13:40 – 14:00 | Approximation of Causal Dependences by Structural Regression Equations Vyacheslav Teterin Russia, The Computer Training Center «Specialist», Yandex School of Data Analysis |
| 14:00 – 14:40 | Toward Causal Machine Learning Prof. Bernhard Schölkopf Germany, Max Planck Institute for Intelligent Systems |
Quantum Learning | |
| 14:40 – 15:00 | Inferring Causal Structure from Statistical Dependences among n Variables — Differences between Classical and Quantum Communication Scenarios Dr. Dominik Janzing Germany, Max Planck Institute for Intelligent Systems |
| 15:00 – 15:15 | Break |
| 15:15 – 15:35 | What Can We Expect from Quantum Machine Learning? Dr. Peter Wittek Spain, ICFO-The Institute of Photonic Sciences, Barcelona |
Submodularity and Learning | |
| 15:35 – 16:15 | Learning with Submodular Functions: A Convex Optimization Perspective Prof. Francis Bach France, Ecole Normale Supérieure de Paris |
| 16:15 – 16:35 | Summarizing Massive Data Sets via Large-Scale Submodular Maximization Prof. Andreas Krause Switzerland, ETH Zürich |
| 16:35 – 16:40 | Flash Presentation with slides for the Poster: Minimizing Submodular Discrete Energies by Integer Re-parameterizations Dr. Tomas Werner Czech Republic, Czech Technical University Prof. Michail Schlesinger Ukraine, International Research and Training Centre of Information Technologies and Systems, National Academy of Science of Ukraine, Cybernetica Centre |
| 16:40 – 16:50 | Break |
Clustering and Sparsity | |
| 16:50 – 17:10 | K-means and Anomalous Clustering Prof. Boris Mirkin Russia, National Research University Higher School of Economics in Moscow, UK, Birkbeck University of London |
| 17:10 – 17:30 | Distributed Coordinate Descent for Regularized Logistic Regression Ilya Trofimov Russia, Yandex Data Factory |
| 17:30 – 17:35 | Flash Presentation with slides for the Poster: Efficient Elastic Net Regularization for Sparse Linear Models in the Multilabel Setting Zachary Chase Lipton ML scientist, Amazon, University of California, San Diego, USA |
| 17:35 – 17:55 | Large-Scale Learning Dr. Mikhail Bilenko USA, Microsoft |
| 17:55 – 18:05 | Break |
| 18:05 – 20:35 | Discussion "Deep Learning". Moderator Dr. Li Deng (expected participants: Pierre Baldi, John Shawe-Taylor, Nathan Intrator, Alexandre Klementiev, Dmitrij Schlesinger and others) |
October 6
Session III
Chair Dr. Andrey Ustyuzhanin, RussiaYandex, Yandex School of Data Analysis
Particular Problems | |
| 9:00 – 9:20 | Zero-Shot Learning Dr. Timothy Hospedales UK, School of Electronic Engineering & Computer Science, Queen Mary University of London |
| 9:20 – 9:40 | A Potential Surface Underlying Meaning? Prof. Sándor Darányi Sweden, Högskolan i Borås, Swedish School of Library and Information Science |
| 9:40 – 10:20 | Can Machine Learning replace Signal Processing? Prof. Nathan Intrator Israel, Blavatnik School of Computer Science, Sagol School of Neuroscience |
| 10:20 – 10:35 | Break |
Generalizations and Physics | |
| 10:35 – 10:55 | Relationship Graph Filtering Drawing on the Example of the Site Hierarchy Extraction Boris Belyaev Russia, Yandex |
| 10:55 – 11:00 | Flash Presentation with slides for the Poster: Modified Naive Bayes with Hurst Exponent as Quantitative Measure of Data Mutual Dependence Gleb Turkanov Russia, Russian Academy of Sciences, Moscow Institute of Physics and Technology |
| 11:00 – 11:05 | Flash Presentation with slides for the Poster: Yandex + CERN: Three Years of Collaboration of Data Science with Particle Physics Dr. Andrey Ustyuzhanin Yandex, Yandex School of Data Analysis |
| 11:05 – 11:10 | Flash Presentation with slides for the Poster: Machine Learning and Optimization of LHC Real-Time Event Stream Filter for New Physics Discoveries Tatiana Likhomanenko Russia, National Research University Higher School of Economics, Yandex Data Factory Dr. Andrey Ustyuzhanin Russia, Yandex, Yandex School of Data Analysis |
Applications | |
| 11:10 – 11:15 | Flash Presentation with slides for the Poster: The Use of Random Forest to Improve Molecular Docking Tools for Drug Design Dr. Pedro J. Ballester France, Cancer Research Centre of Marseille |
| 11:15 – 11:20 | Flash Presentation with slides for the Poster: Bank Failure Prediction Using Hybrid Classifier Ensembles of Random Sub-Spaces and Bagging Assoc. Prof. Halil İbrahim Erdal, Dr. Aykut Ekinci Turkey, Turkish Cooperation and Coordination Agency, F. A. Hayek Visiting Scholar, George Mason University |
| 11:20 – 11:25 | Flash Presentation with slides for the Poster: Trading the FTSE100 Index – 'Adaptive' Modeling and Optimisation Techniques Dr. Andreas Karathanasopoulos Associate Professor: American University of Beirut |
| 11:25 – 11:30 | Flash Presentation with slides for the Poster: Kernel-based Machine Learning from Multiple Information Sources: Learning Theory, Algorithms, and Applications in Visual Image Recognition and Computational Biology Prof. Marius Kloft Germany, Humboldt University of Berlin Department of Computer Science |
| 11:30 – 11:35 | Flash Presentation with slides for the Poster: Bots Filtering in Crowd-Sourcing Dr. Andrey Mishchenko Russia, Yandex |
| 11:35 – 11:40 | Flash Presentation with slides for the Poster: Yandex Recommender System Dr. Michael Roizner Russia, Yandex |
| 11:40 – 11:45 | Flash Presentation with slides for the Poster: Internet Search Trends Extraction: Using Topic Modeling Ideas for Clustering User Queries Denis Elshin, Daria Gubar Russia, Yandex |
| 11:45 – 11:50 | Flash Presentation with slides for the Poster: Expansion of Video Categories Through Link and Users Graphs Boris Okun Russia, Yandex |
| 11:50 – 11:55 | Flash Presentation with slides for the Poster: Deep Neural Networks for Fast Object Detection Anton Slesarev Russia, Yandex |
| 11:55 – 12:10 | Break |
| 12:10 – 12:30 | MatrixNet Applications at Yandex Michael Levin Russia, Yandex Data Factory, Yandex |
A New View on Fundamental Problems | |
| 12:30 – 12:50 | Implicit Modeling — A Generalization of Discriminative and Generative Approaches Dr. Dmitrij Schlesinger Germany, Dresden University of Technology |
| 12:50 – 13:10 | Online Learning or Don’t Look Ahead Andrey Gulin Russia, Yandex |
| 13:10 – 14:00 | Lunch |
Session IV
Chair Prof. Alexander Gammerman, UKRoyal Holloway, University of London
| 14:00 – 15:20 | Intelligent Learning: Similarity Control and Knowledge Transfer Prof. Vladimir Vapnik USA, Columbia University, Facebook |
| 15:20 – 15:35 | Break |
| 15:35 – 19:00 | Social program |
| 19:00 | Banquet |
October 7
Session V
Chair Dr. Evgeny Burnaev, RussiaInstitute for Information Transmission Problems of Russian Academy of Sciences
Machine Learning Foundations | |
| 9:00 – 9:40 | Large-Scale Probabilistic Prediction With and Without Validity Guarantees Prof. Vladimir Vovk UK, Royal Holloway, University of London |
| 9:40 – 10:20 | Minimax Deviation Strategies for Machine Learning and Recognition with Short Learning Samples Prof. Michail Schlesinger, Evgeniy Vodolazskiy Ukraine, International Research and Training Centre of Information Technologies and Systems, National Academy of Science of Ukraine, Cybernetica Centre |
| 10:20 – 10:35 | Break |
| 10:35 – 10:40 | Flash Presentation with slides for the Poster: Global and Local Complexity Measures for Transductive Learning Ilya Tolstikhin Germany, Max Planck Institute for Intelligent Systems, Department of Empirical Inference |
| 10:40 – 10:45 | Flash Presentation with slides for the Poster: Explaining Predictions of Arbitrary Model with Feature Contributions Prof. Igor Kononenko Slovenia, Laboratory for Cognitive Modeling University of Ljubljana |
| 10:45 – 11:05 | Multidimensional Conditional Probability Estimation Using the V-Matrix Method Dr. Rauf Izmailov USA, Vencore Labs Prof. Vladimir Vapnik USA, Columbia University, Facebook |
Combinatorics | |
| 11:05 – 11:25 | Is It Really NP-Hard to Survive in Big-Data? Prof. Vadim E. Levit Israel, Department of Computer Science and Mathematics, Ariel University |
| 11:25 – 11:45 | Committee Polyhedral Separability: Complexity and Polynomial Approximation Prof. Michael Khachay Russia, Krasovsky Institute of Mathematics and Mechanics, Ural Federal University, Ekaterinburg |
Text Analysis and Understanding | |
| 11:45 – 11:50 | Flash Presentation with slides for the Poster: Functional Analysis of F0 Contours for Recognition of Paralinguistics from Speech Dmytro Prylipko Germany, Yandex |
| 11:50 – 11:55 | Flash Presentation with slides for the Poster: Faster Recurrent Neural Network Language Modeling Toolkit with Logarithmic and Sub-logarithmic Cost Function Approximation Algorithms Anton Bakhtin Russia, Yandex |
| 11:55 – 12:00 | lash Presentation with slides for the Poster: Linguistic Regularization of Topic Models Anna Potapenko Russia, Yandex School of Data Analysys Prof. Konstantin Vorontsov Russia, Yandex |
| 12:00 – 12:15 | Break |
| 12:15 – 12:55 | Thinking on your Feet: Reinforcement Learning for Incremental Language Tasks Prof. Jordan Boyd-Graber USA, University of Colorado |
| 12:55 – 13:15 | Inducing Crosslingual Distributed Representations of Words Dr. Alexandre Klementiev Germany, Amazon |
| 13:15 – 14:05 | Lunch |
Session VI
Chair Dr. Uri Degen, IsraеlTSG IT Advanced Systems Ltd.
| 14:05 – 14:45 | Additive Regularization of Topic Models: Towards Exploratory Search and Other Multi-Criteria Applications Prof. Konstantin Vorontsov Russia, Yandex |
Applications | |
| 14:45 – 15:25 | Detecting of hidden patterns in air/maritime/human situation picture Dr. Uri Degen, Dr. Leonid Shvartser, Prof. Igor Korotayev, Dr. Rima Gandlin, Alexander Knafel Israеl, TSG IT Advanced Systems Ltd. |
| 15:25 – 15:45 | Identification of Activity Patterns in Communication Data without Content Dr. Uri Degen, Dr. Leonid Shvartser, Shmuel Teppler Israеl, TSG IT Advanced Systems Ltd. |
| 15:45 – 16:25 | Reliable Diagnostics by Conformal Predictors Prof. Alexander Gammerman UK, Royal Holloway, University of London |
| 16:25 – 16:40 | Break |
| 16:40 – 17:00 | Dynamic Style Analysis of Hedge Funds and Kalman Smoother Dr. Leonid Shvartser Israеl, TSG Advanced Systems Ltd Michael Markov USA, MPI International Inc Dr. Olga Krasotkina Russia, Moscow State University Prof. Vadim Mottl' Russia, Computing Center of the Russian Academy of Sciences |
| 17:00 – 17:05 | Flash Presentation with slides for the Poster: The Machine Learning Techniques in Seismic Data Processing and Interpretation Problems Maxim Ryabinskiy, Dr. Dmitriy Finikov Russia, Yandex.Terra |
| 17:05 – 17:25 | Sparse Regression and Data Preparation for Machine Learning in Seismic Data Processing Maxim Ryabinskiy, Dr. Dmitriy Finikov, Dr. Nikita Zelinsky Russia, Yandex.Terra |
| 17:25 – 17:30 | Flash Presentation with slides for the Poster: Deep Learning for Image Retrieval Artem Babenko Russia, Yandex |
| 17:30 – 17:35 | Flash Presentation with slides for the Poster: Query Model for Image Search based on User Clicks and NN Features Dmitry Krivokon, Alexey Gorodilov Russia, Yandex |
Current Methodological Progress | |
| 17:35 – 17:40 | Flash Presentation with slides for the Poster: Conformalized Kernel Ridge Regression and Its Efficiency Dr. Evgeny Burnaev Russia, Institute for Information Transmission Problems of Russian Academy of Sciences Prof. Vladimir Vovk UK, Royal Holloway, University of London |
| 17:40 – 17:45 | Flash Presentation with slides for the Poster: Locally Isometric and Conformal Low-dimensional Data Representation in Data Analysis Prof. Alex Bernshtein, A.P. Kuleshov, Yu.A. Yanovich Russia, National Research University Higher School of Economics |
| 17:45 – 17:50 | Flash Presentation with slides for the Poster: Tensor perspective of deep neural networks Prof. Dmitry Vetrov National Research University Higher School of Economics, Yandex School of Data Analysis, Skoltech Alexander Novikov, Dmitry Podoprikhin Skoltech Anton Osokin INRIA - SIERRA project-team |
| 17:50 – 18:00 | Break |
| 18:00 – 20:00 | Discussion "What Learning Problem is About: Brute Force and Intelligent Paradigms of Learning". Moderator: Vladimir Vapnik (expected participants: Konstantin Vorontsov, Anatoli Michalski, Vladimir Vovk, Bernhard Schölkopf, Varvara Tsurko, Andrey Ustyuzhanin and others). |
| 20:00 – 20:30 | Closing Remarks Arkady Volozh Russia, Yandex Prof. Elena Bunina Russia, Yandex, Yandex School of Data Analysis and Moscow State University |
October 8
| 9:00 – 15:00 | Business session «Machine Learning and Big Data: Business Challenges» |
| 15:00 | Conference closes. Networking |
Key speakers
Francis Bach, France
Pierre Baldi, USA
Li Deng, USA
Alexander Gammerman, UK
Nathan Intrator, Israel
Vadim Levit, Israel
Bernhard Schölkopf, Germany
Vladimir Vapnik, USA
Vladimir Vovk, UK
Lior Wolf, Israel
Program Committee
Maxim Babenko
Jordan Boyd-Graber
Elena Bunina
Alexander Gammerman
Boris Mirkin
Ilya Muchnik
Andrey Raigorodsky
Konstantin Rudakov
Bernhard Schölkopf
Evgeny Shchepin
Albert Shiryaev
Konstantin Vorontsov
Organising Committee
Elena Bunina
Stanislav Fedotov
Evgenia Kulikova
Nataliya Makarova
Ilya Muchnik
Elena Shiryaeva
Arkady Volozh