somewhere in the Russian forest (learning advanced techniques of data analysis)

Marina Dubova

I am a Cognitive Science PhD student at Indiana University. My current research revolves around two major themes:

  1. I study grounded language learning in humans and AI, focusing on the effects of various aspects of naturalistic data, tasks, and training regimes on language acquisition. I take inspiration from the idea that natural communication is grounded in one’s goals, perception, and action in the world. I examine 1) why grounding is essential for human language learning, 2) what concrete aspects of grounding should be incorporated into language-learning AI systems to improve their capabilities, and 3) how this can be done.

  2. I study the mechanisms underlying the development of communication systems (e.g. human languages!) in groups of agents. Here, I am interested in the questions of 1) how the properties of evolved communicative systems depend on the learning mechanisms, communicative goals, and the structure of social interactions between the agents 2) which pressures lead a group to converge on a shared set of communicative conventions 3) which pressures force a group to increase and diversify their set of communicative conventions (e.g. number of categories that words refer to).

I combine theoretical analyses, behavioral experiments, and computational simulations in my research. I also like to think about scientific progress in general, which motivates me to constantly improve how I conduct my research.

I am always enthusiastic to discuss your/my research, ideas, and the fundamental questions of mind and development! Feel free to contact me :)

Cognitive Scientists who shaped my thinking the most: James McClelland & David Rumelhart, Andy Clark, Linda Smith, Sam Gershman, Rob Goldstone, Pierre-Yves Oudeyer, Randall Beer, Uri Hasson, James DiCarlo, Luc Steels, Justin Wood

I am grateful to be supported by these people: Arseny Moskvichev, Polina Greskova, Alexey Filippov, Roman Tikhonov, Vili Hätönen, George Moroz, Grigory Anufriev, Thomas Gorman, Eleanor Schille-Hudson, Rob Goldstone, Peter Todd + many others

News

Sept 17 - Sept 25: I am presenting our work "The Influences of Category Learning on Perceptual Reconstructions" at the 61st Annual Meeting of the Psychonomic Society: check out our poster here!

Sept 17 - Sept 25: Presenting our work "Reinforcement Communication Learning in Different Social Network Structures" at NetSci!

July 29 - Aug 1: Participating at CogSci annual conference!

July 18: Presenting our work "Reinforcement Communication Learning in Different Social Network Structures" at ICML Language and Reinforcement Learning workshop!

July 16: I'm presenting the work "Effects of Supervision, Population Size, and Self-Play on Multi-Agent Reinforcement Learning to Communicate" at ALIFE conference (feel free to join the discussion!)

July 13 - July 18: I am participating at ALIFE and ICML conferences!

March 12 .... 2020: I am isolating at home, drinking tea, and working *very productively*

June 22July 3, 2020: I will attend SINSA2020 (won fee remission by IUNI-SINSA Scholars program)

June 14 – July 10, 2020: I will attend the Complex Systems Summer School at Santa Fe Institute (whoa!)

March 1 – March 3, 2020: I gave a talk on the mechanisms of categorical perception at Emory University, Georgia (Mechanisms of Learning Forum)