Fall 2016

September 14, 2016

Tight Bounds for Multi-Reference Alignment

Speaker: Jonathan Weed (Math, MIT)

How should one estimate a signal, given only access to noisy versions of the signal corrupted by unknown circular shifts? This simple problem has surprisingly broad applications, in fields from structural biology to aircraft radar imaging....

September 21, 2016

Delay, Memory, and Messaging Tradeoffs in Distributed Service Systems

Speaker: Martin Zubeldia (LIDS)

We consider the following distributed service model: jobs with independent processing times of unit mean arrive as a Poisson process of rate \lambda N, with 0<\lambda<1 fixed, and are immediately dispatched to one of several queues...

September 28, 2016

Neural Auto-associative Memory Via Sparse Recovery

Speaker: Ankit Rawat (RLE)

An associative memory is a framework of content-addressable memory that  stores a collection of message vectors (or a data set) over a neural network while enabling a neurally feasible mechanism to recover any message in the data set from...

October 5, 2016

Bayesian Heuristics for Group Decisions

Speaker: Amin Rahimian (Electrical & Systems Engineering, University of Pennsylvania)

We propose a model of inference and heuristic decision making in groups that is rooted in the Bayes rule but avoids the complexities of rational inference in partially observed environments with incomplete information. According to our model...

October 12, 2016

An Introduction to Sparse PCA

Speaker: Jerry Li (CSAIL)

A very powerful tool for dimension reduction in many applications is principal component analysis (PCA), which really just means to project your data matrix onto its most significant directions, i.e. its top singular vectors, and hope those...

October 19, 2016

On Nonlinear Shaping Filters with Minimal Output Peak Value

Speaker: Omer Tanovic (LIDS)

One of the main challenges in modern digital communication systems is high peak-to-average power ratio (PAPR) of signals being transmitted. For example, this is the major drawback of orthogonal frequency division multiplexing (OFDM), a...

October 26, 2016

Asymmetric Numeral Systems: a new approach to entropy coding

Speaker: Jennifer Tang (LIDS)

When it comes to data compression, having a small sized compressed file is important, but as people continue to create and share more and more data, having a method that can achieve this compression rate quickly and efficiently is also...

November 2, 2016

Ignoring Reality to Improve Approximation / Competitive Ratios in Expectation

Speaker: Will Ma (ORC)

We consider two stochastic optimization algorithms which make adaptive decisions over time as information is revealed, except instead of making decisions based on the realized current state, they consider the distribution of all possible...

November 9, 2016

Learning a tree-structured Ising model in order to make predictions

Speaker: Mina Karzand (RLE)

A primary reason for the widespread use of graphical models is the existence of efficient algorithms for prediction from partial observations. But it's not clear how much data is required to learn a model so that such predictions are...

November 16, 2016

Variational inference via measure transport: algorithms for Bayesian filtering and smoothing

Speaker: Alessio Spantini (AeroAstro, MIT)

We describe a nonparametric variational inference method that approximates an intractable target measure as the pushforward of a tractable distribution (e.g., a Gaussian) through a transport map, and that can approximate arbitrary...

November 23, 2016

Where did I leave my self-driving car? A Mathematical Perspective on Robot Localization and Mapping

Speaker: Luca Carlone (LIDS)

Localization and mapping are the backbone of many robotics applications including self-driving cars and autonomous drones. In this talk I discuss a Maximum Likelihood formulation of robot Simultaneous Localization and Mapping (SLAM). While...

November 30, 2016

LIDS & Stats Tea - George Chen [CANCELLED]

Speaker: George Chen (CSAIL)

December 7, 2016

Some Thoughts About the Most Informative Boolean Function Conjecture

Speaker: Or Ordentlich (LIDS)

Suppose X is a uniformly distributed n-dimensional binary vector and Y is obtained by passing X through a binary symmetric channel with crossover probability \alpha. A recent conjecture by Courtade and Kumar postulates that I(f(X);Y)\leq 1-h...