Learning - overview

The Learning PIC covers all aspects of machine learning from learning theory, development of novel learning algorithms to applications of learning. Learning often plays a key role in other research areas such as automated reasoning, computational biology, perception, etc. [link to other PICs]

The key academic conferences for the Learning PIC are ICML, KDD, NIPS, UAI, ICDM and the machine learning tracks at confereces such as AAAI and IJCAI.

The key academic journals would be JMLR, MLJ, DMKD, KAIS and TKDD.

Key accomplishments from IBM Research in this area include:
+ "Checkers Player," Arthur Samuel, 1959
+ "TD-Gammon - Computer Backgammon," Gerald Tesauro, 1992
+ "Watson Jeopardy Challenge," David Ferrucci et. al., 2011

Learning is a key component in various IBM research projects, including:
+ Debater
+ Medical Sieve

Recent Conference Activity

Irina Rish, Invited Speaker at NIPS 2016


Recent Paper Awards

Tsuyoshi Ide and Amit Dhurandhar. Informative Prediction based on Ordinal Questionnaire Data. IEEE Intl. Conference on Data Mining (ICDM), 2015 (Best paper candidate)

Amit Dhurandhar, Rajesh Ravi, Bruce Graves, Gopikrishnan Maniachari and Markus Ettl. Robust System for Identifying Procurement Fraud. Assoc. for Adv. in Artificial Intelligence (AAAI), 2015. (Deployed Application Award)

Publications at NIPS 2015:

Closed-form Estimators for High-dimensional Generalized Linear Models
Eunho Yang, IBM Research; Aurelie Lozano, IBM Research; Pradeep Ravikumar, University of Texas at Austin

Parallel Recursive Best-First AND/OR Search for Exact MAP Inference in Graphical Models
Akihiro Kishimoto, IBM Research; Radu Marinescu, IBM Research; Adi Botea, IBM Research

Backpropagation for Energy-Efficient Neuromorphic Computing
Steve Esser, IBM Research; Rathinakumar Appuswamy, IBM Research; Paul Merolla, IBM Research; John Arthur, IBM Research; Dharmendra Modha, IBM Research

Learning with Group Invariant Features: A Kernel Perspective.
Youssef Mroueh, IBM; Stephen Voinea, MIT; Tomaso Poggio, MIT

Robust Gaussian Graphical Modeling with the Trimmed Graphical Lasso
Eunho Yang, IBM Research; Aurelie Lozano, IBM Research

Information-theoretic lower bounds for convex optimization with erroneous oracles
Yaron Singer, Harvard University; Jan Vondrak, IBM Research

Publications at AAAI 2016:

Selecting Near-Optimal Learners via Incremental Data Allocation
Ashish Sabharwal, Horst Samulowitz, Gerry Tesauro


Publications at Machine Learning Journal 2015:

Improving Classification Performance through Selective Instance Completion

Amit Dhurandhar and Karthik Sankarnarayanan


Publications at UAI 2014:

Structured Proportional Jump Processes

   Tal El-Hay, Omer Weissbrod, Elad Eban, IBM Research; Maurizio Zazzi, Univeristy of Siena;
   Francesca Incardona, EuResist Network GEIE




Related links

Professional Interest Communities at IBM Research

Learning is one of five interrelated research areas under the Cognitive Computing umbrella:





Cognitive Computing

Algorithms and Theory

IBM Research Accomplishments