David Alvarez Melis
I'm a graduate student in the EECS department at MIT,
working with Tommi
Jaakkola. I'm broadly interested in Machine Learning and
Natural Language Processing, particuarly in settings involving structured data. Some topics that I am currently (or have been) working on:
 Perturbationbased approaches to interpretability in machine learning
 Optimal transport
 Generative adversarial networks
 [Unsemi]supervised domain adaptation and machine translation.
 Tree and sequence embedding and generation.
 Semantic representations, their manifold structure and geometric properties.
 Optimization: Semidefinite programming, MMW, convex optimization.
Before coming to MIT, I spent a year at IBM's T.J. Watson
Research Center, working with Ken Church and others in the Speech
Recognition Group.
Even before that, I completed a BSc in Applied Math at ITAM
(Mexico City) where I was advised by Carlos
Bosch and an MS in Math at the Courant
Institute, NYU, where I was advised by Mehryar Mohri.
MIT Computer Science and Artificial Intelligence Lab
Stata Center, Bldg 32G496
Cambridge, MA 02139
Email: d_alv_mel_[at]_mit_[dot]_edu (humans: remove underscores).
LinkedIn
Google Scholar
Research

Structured Optimal Transport.
David AlvarezMelis, Tommi S. Jaakkola, Stefanie Jegelka.
NIPS Workshop on Optimal Transport for Machine Learning, 2017. Extended Oral.
arXiv (coming soon)

Distributional Adversarial Networks.
Chengtao Li, David AlvarezMelis, Keyulu Xu, Stefanie Jegelka, Suvrit Sra.
Preprint, 2017.
arXiv
code

A causal framework for explaining the predictions of blackbox sequencetosequence models.
David AlvarezMelis, Tommi Jaakkola.
EMNLP 2017.
.pdf from ACL
arXiv (w/supplementary)
MIT News
.bib

Treestructured decoding with doublyrecurrent neural networks.
David AlvarezMelis, Tommi Jaakkola.
ICLR 2017.
.pdf
code
.bib

Word Embeddings as Metric Recovery in Semantic Spaces.
Tatsunori B. Hashimoto, David AlvarezMelis, Tommi Jaakkola.
TACL 2016, presented at ACL 2016.
.pdf
.bib

Topic Modeling in Twitter: aggregating tweets by conversations.
David AlvarezMelis, Martin Saveski.
ICWSM 2016.
.pdf
.bib

Word, graph and manifold embedding from Markov processes.
Tatsunori B. Hashimoto, David AlvarezMelis, Tommi Jaakkola.
NIPS 2015 Workshop on Nonparametric Methods for Large Scale Representation Learning. Oral presentation.
Preprint
arXiv

A translation of "The characteristic function of a random phenomenon" by Bruno de Finetti.
David AlvarezMelis, Tamara Broderick
arXiv

The Matrix Multiplicative Weights Algorithm for Domain Adaptation.
David AlvarezMelis (advisor: Mehryar Mohri).
MS Thesis, Courant Institute 2013.
.pdf

LaxMilgram's Theorem: Generalizations and Applications.
David AlvarezMelis (advisor: Carlos Bosch Giral).
BSc Thesis, ITAM 2011. (In Spanish)
.pdf
Projects
Some course projects:

Distance Metric Learning Through Convex Optimization
Course: Numerical Optimization (NYU), taught by Margaret Wright.
.pdf

Nests and Tootsie Pops: Bayesian Sampling with Monte Carlo.
with Michael Khanarian.
Course: Monte Carlo Methods (NYU), taught by Jonathan Goodman.
.pdf

A Weighted Finite State Machine Implementation of Alignment and Translation Models
with Andres MunozMedina
Course: Speech Recognition (NYU), taught by Mehryar Mohri.
.pdf
slides

Sentiment Classification in Twitter: A Comparison between Domain Adaptation and Distant Supervision
with Michael Khanarian.
Course: Statistical Natural Language Processing (NYU), taught by Slav Petrov.
.pdf
Teaching
Explaining is understanding.
The following extract is from David Goldstein's book on Feynman:
"Feynman was a truly great teacher. He prided himself on being able to devise ways to explain even the most profound ideas to beginning students. Once, I said to him, "Dick, explain to me, so that I can understand it, why spin onehalf particles obey FermiDirac statistics." Sizing up his audience perfectly, Feynman said, "I'll prepare a freshman lecture on it." But he came back a few days later to say, "I couldn't do it. I couldn't reduce it to the freshman level. That means we don't really understand it."
Some current and past courses I have TA'd:
At MIT:
At NYU: