Humans' Models and Learned Models

par

Jason Eisner

Computer Science at Johns Hopkins University

Vendredi 21 septembre 2018, 10:00-12:00, Salle 3195, Pavillon André-Aisenstadt

La conférence sera présentée en anglais

Résumé:

Humans like to construct intricate theories of the world, often involving representations
of hidden processes. AI systems that are aware of these theories and representations can use them
(a) to reason about novel situations and (b) to better collaborate with humans. Typically, a theory
of some domain can be made precise as a structured generative probability model, with a large set
of parameters to be filled in through machine learning.

This talk will consider some diverse examples of rich probabilistic modeling, drawn from natural
language processing and other human-computer collaboration settings. It will sketch how novel ML
algorithms were needed to handle the challenges of efficient inference and learning in these
models. By way of closing, we will discuss how to accelerate this ML4AI development process, via a
common layer of declarative infrastructure that can be equipped with generic algorithms.

Bio :

Jason Eisner is Professor of Computer Science at Johns Hopkins University, where he is also affiliated with the Center for Language and Speech Processing, the Machine Learning Group, the Cognitive Science Department, and the national Center of Excellence in Human Language Technology. He is an action editor for JAIR and TACL. His goal is to develop the probabilistic modeling, inference, and learning techniques needed for a unified model of all kinds of linguistic structure. His 125+ papers have presented various algorithms for parsing, machine translation, and weighted finite-state machines; formalizations, algorithms, theorems, and empirical results in computational phonology; and unsupervised or semi-supervised learning methods for syntax, morphology, and word-sense disambiguation. He is also the lead designer of Dyna, a new declarative programming language that provides an infrastructure for AI research. He has received two school-wide awards for excellence in teaching.