Microsoft Research

It is quite easy to believe that the recently proposed approach to machine translation, called neural machine translation, is simply yet another approach to statistical machine translation.

The conventional semiconductor and electronics assembly industry has relied on ever increasing densities of integration from high capital cost manufacturing and disparate supply chains and assembly

Transition-based models use a transition system, or abstract state machine, to model structured prediction problems, for example syntactic dependency parsing, as a sequence of actions.

To accelerate the development of more effective cancer therapies, we have developed and optimized a cancer drug target discovery platform that entails high throughput siRNA and focused small

We study the problem of allocating a set of indivisible items among agents with additive valuations with the goal of maximizing the geometric mean of the agents' valuations, i.e., the Nash social

In my vision for the future, wearable computing devices will interpret rich visual real-world environments in real-time, providing services to assist in our daily lives with Continuous

In my Ph.D. thesis, I propose mechanisms to address the inefficiency caused by existing memory abstractions, which manage and access memory at different granularities at different resources.

With the advent of cloud storage and computing, adopting cloud services to reduce operational, maintenance and storage costs, is becoming increasingly common --- both for individuals and corporations.

Due to the complexity of the problems, collaboration of people with different backgrounds and expertise is becoming increasingly necessary.

Neural networks have become the main workhorse of supervised learning, and their efficient training is an important technical challenge which has received a lot of attention.

We proposed a doubly stochastic primal-dual coordinate optimization algorithm for regularized empirical risk minimization that can be formulated as a saddle-point problem using convex conjugate

When a bug is found in a long-running distributed system, developers typically start by identifying (i) which events in the execution caused their system to arrive at the unsafe state, and (ii)

Natural images exhibit high information correlation across pixels. Bilateral filtering provide a simple yet powerful framework for information propagation across pixels.

This talk has two inter-twined aims.

We present a new non-malleable commitment protocol. Our protocol has the following features: 1) The protocol has only three rounds of interaction.

Opening Keynote (Lunch talk): Bobby Kleinberg - Incentivizing Exploration We explore the misalignment of incentives in systems that use crowdsourced effort to solve large problems made up of smaller

On October 13th 1972, a Uruguayan Air Force plane was lost without a trace in the Andes.

Transaction processing is a fundamental aspect of database design, and concurrency control has remained a key challenge in high performance transaction management.

Ishai, Kushilevitz, Ostrovsky and Sahai (STOC`07, SIAM JoC 2009) introduced the powerful ΓÇ£MPC-in-the-headΓÇ¥ technique that provided a general transformation of information-theoretic MPC protocols

Pony is an actor-model, capabilities-secure, native programming language.

In the setting of cloud computing a user wishes to delegate its data, as well as computations over this data, to a cloud provider.

In this talk I provide an overview of information-theoretic bounded-rationality for planning in sequential decision problems.

Many problems in real-world applications involve predicting several random variables which are statistically related.

Interactive proofs have had a dramatic impact on Complexity Theory and Cryptography.

The shift to multi-core architectures in the past ten years pushed developers to write concurrent software to leverage hardware parallelism.

N-gram models are essential in any kind of text processing; they offer simple baselines that are surprisingly competitive with more complicated state

Miklos Racz - Sequence assembly from corrupted shotgun reads Reconstructing a sequence from its short substrings, called reads, is a central problem in DNA sequencing.

Real-valued vector representations of words (i.e. embeddings) that are trained on naturally occurring data by optimising general-purpose objectives are useful for a range of downstream language tasks.

Many real-world problems require making decisions that involve multiple possibly conflicting objectives.

Gireeja Ranade - A tiger by the tail: When multiplicative noise stymies control Consider an unstable discrete-time scalar linear system that is observed over a channel corrupted by continuous