Reprezentáció tanulás elmeletének kifejlesztése az indukciós biasok hatásainak szimulálásán keresztül

Nyomtatóbarát változatNyomtatóbarát változat
Cím angolul: 
Developing a theory of representation learning by simulating the effect of inductive biases
BSc szakdolgozat téma - fizikus
MSc diplomamunka téma - nanotechnológia és anyagtudomány
MSc diplomamunka téma - optika és fotonika
MSc diplomamunka téma - kutatófizikus
MSc diplomamunka téma - nukleáris technika
MSc diplomamunka téma - orvosi fizika
Fiser Jozsef
Email cím:
CEU, Kognitív Tudományi Tanszék
egyetemi tanár
Varga Imre
Email cím:
Fizikai Intézet, Elméleti Fizika Tanszék
egyetemi docens
Strong interest in understanding the algorithmic structure of learning problems, ability to independently process literature in English and implement software in python, basic knowledge of linear algebra, calculus and probability theory, experience with machine learning algorithms is a plus
Representation learning is a process that an agent uses to alter which pieces of information it retains from its input stream, and which pieces it discards, implementing an intelligent compression scheme. There are strong theoretical underpinnings for using Bayesian inference for compression problems, and also for learning to make good decisions based on a pre-existing representation, but there is currently no theory describing how representations should change over time in a task-based setting, which would be of major interest to both cognitive science and machine learning. We want to develop such a theory by formalising the interplay between the inductive biases of the agent, the constraints on available resources and the task objective in a learning algorithm. The student will participate in the algorithmic design of agents and tasks testing the theory, the implementation of the tree-search based and gradient-based algorithmic components, and the evaluation of the results.


Hozzáférés nincs korlátozva