Research talks

Learning parameters of causal generative models by the approximate method of moments.
Statistical models for natural sounds. The sounds for this talk can be found here.
Auditory Gestalt from natural scene-statistics.
Bayesian Amplitude Demodulation: A new family of statistical signal processing algorithms.
A new generative model for sounds: The Gaussian Modulation cascade process Sounds drawn from the forward model: sound1 and sound2
A Maximum Likelihood Algorithm For Slow Feature Analysis, the demos shown in the talk can be found here. Load up annealing_dem.m - you'll need the .mat files listed in the script too
Online inference of a stimulus with a dynamically changing variance: modeling the data of Fairhall et al (2001, Nature) using a factorial linear dynamical system here. (a summary of Fairhall and Smirnakis's results is here)