Personal tools
You are here: Home Events ANC Workshop: Militos Allamanis and Zhanxing Zhu, Chair: Chris Williams

ANC Workshop: Militos Allamanis and Zhanxing Zhu, Chair: Chris Williams

— filed under:

What
  • ANC Workshop Talk
When Oct 06, 2015
from 11:00 AM to 12:00 PM
Where IF 4.31/4.33
Add event to calendar vCal
iCal

Militos Allamanis

"Learning Source Code Representations"

Recent work in representation learning, such as word2vec [1] and paragraph vectors [2], show that powerful representations can be extracted from natural language. In this work, we explore representation learning for source code. To show that the extracted representations capture semantic properties of code, we use the representation of method and class bodies to predict their names.

[1] Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. Efficient Estimation of Word Representations in Vector Space. In Proceedings of Workshop at ICLR, 2013.
[2] Le, Quoc V., and Tomas Mikolov. "Distributed representations of sentences and documents." arXiv preprint arXiv:1405.4053 (2014).

 

Zhanxing Zhu 

"Covariance-Controlled Adaptive Langevin Thermostat for Large-Scale Bayesian Sampling"

 Abstract: Monte Carlo sampling for Bayesian posterior inference is a common approach used in machine learning. Often the Markov Chain Monte Carlo procedures that are used are discrete-time analogues of associated stochastic differential equations (SDEs). These SDEs are guaranteed to leave invariant the required posterior distribution.

 One area of current research addresses the computational benefits of stochastic gradient methods in this setting. Existing techniques rely on estimating the variance or covariance of the subsampling error, and typically assume constant variance. In this talk, I will talk about a covariance-controlled adaptive Langevin thermostat that can effectively dissipate parameter-dependent noise while maintaining a desired target distribution. The described method achieves a substantial speedup over popular alternative schemes for large-scale machine learning applications.