Microsoft Research

Microsoft Research – 12 августа 2016, 2:21

Match the applications to the theorems: (i) Find the variance of traffic volumes in a large network presented as streaming data.

I will try to explain some key principles in modern mathematics which combine ideas from combinatorics and probability.

We describe an approach for synthesizing data representations for concurrent programs.

We present Concurrit, a novel testing technique and domain-specific language (DSL) for unit and system testing of concurrent programs.

Planning under uncertainty, typically modeled with Markov Decision Processes (MDPs), has always been considered a core skill of an intelligent agent.

1:00-3:00: Applications (UIUC, UCB, 60 minutes each) John Hart (UIUC) Avascholar (20 mins) Minh Do (UIUC ) 3D-Reconstruction (20 mins) Gerald Friedman (UCB): ΓÇ£PyCASP: Scalable Multimedia Content

SPEAKER: Francois Rodier TITLE: Asymptotic nonlinearity of Boolean functions ABSTRACT: The nonlinearity of Boolean functions on the space Fm2 is important in cryptography.

In this talk, we present some of our recent attempts on tagging visual data with textual descriptions. Annotation process learns to predict the tags from examples.

Starting from the design of robust Hopfield works in the early 1990s to SVMs to structured output problems, a variety of methods have been proposed for the solution of optimization problems arising

Over the past two decades, two statistical machine learning frameworks, graphical models, and structurally constrained (sparse, low-rank, etc.) statistical models have proved very popular and

In this talk I will discuss two learning problems: 'Learning to Rank from Pairwise Preferences' and 'Clustering from Pairwise Similarity Information'.

Microsoft Research – 12 августа 2016, 2:13

Partially Observable Markov Decision Process is an elegant and general model for planning under uncertainty.

Carnegie Mellon - Peter Scupelli, Bruce Hanington, Priscilla Mok, Loretta Neal, Somya Jampala Umea University - Shivanjali Tomar, Siri Johansson Royal College of Art - James Auger, Neil Usher, Shing

Ranking problems arise in an increasing number of applications, including for example information retrieval, recommendation systems, computational biology, drug discovery, and a variety of

Long after MIMO communication has been proposed as a viable path around the rate limitations of point-to-point communication, achieving its most general forms remains an elusive goal.

The web is a vast repository of human knowledge.

Kidney exchange involves creating a non-monetary marketplace through which incompatible donors and patients can take part in exchanges, so that each patient in the exchange receives a transplant

One of the popular directions for scaling up reinforcement learning algorithms is the use of spatio-temporal abstractions.

Classical machine learning makes the i.i.d. (independently and identically distributed) assumption on the data instances. Many real world problems have inherent relational structure where the i.i.d.

We look at the problem of supervised dimension reduction (SDR) from three perspectives.

Most real-world applications involve massive databases comprising of a large number of entities and n-ary relationships, each of which in turn are associated with multiple attributes.

Motivated from several real world problems we consider the problem of designing SVM classifiers which are robust to uncertainty in the Kernel matrices.

The Data Race Freeness (DRF) property has been advocated as the de-facto technique for reasoning about concurrent programs with a relaxed memory semantics.

Many models for sparse regression typically assume that the covariates are known completely, and without noise. Particularly in high-dimensional applications, this is often not the case.

Partial label scenarios arise commonly in web applicationssuch as hierarchical classification, multi-label classification and informationextraction from web pages.

Kenji Takeda from Microsoft Research chairs this session at Faculty Summit 2012. This session highlights some of the latest developments and applications of functional programming.

Microsoft Research – 12 августа 2016, 2:09

Clustering is the problem of finding a 'good'’’ partition of a set of data points in d space into clusters (groups, each consisting of 'nearby'’’ points). In the worst-case, the problem is hard.