Cím angolul:
Automating doubly Bayesian inference using probabilistic programming languages
Típus:
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
Félév:
2023/24/2.
Témavezető:
Név:
Fiser Jozsef
Email cím:
fiserj@ceu.edu
Intézet/Tanszék/Cégnév:
CEU, Kognitív Tudományi Tanszék
Beosztás:
egyetemi tanár
Konzulens:
Név:
Varga Imre
Email cím:
varga.imre@ttk.bme.hu
Intézet/Tanszék:
Fizikai Intézet, Elméleti Fizika Tanszék
Beosztás:
egyetemi docens
Elvárások:
Strong interest in algorithms for statistical inference (interest in human brain and behaviour is a plus), ability to independently process literature in English, strong programming skills (python is essential, experience with any PPL or deep learning packages is a plus), knowledge of probability theory (experience with Bayesian inference and sampling algorithms is a plus)
Leírás:
Modelling autonomous agents (such as humans) requires us to make inferences about inference algorithms they might run internally to make decisions based on stimuli. Algorithms for such doubly Bayesian inference problems have been dubbed cognitive tomography, and underlie a set of recent results in cognitive science. In order to accelerate the development of such algorithms, we want to make use of probabilistic programming languages (PPL) that allow us to run inference by just specifying the statistical model. The student will participate in specifying a class of models for which automation is the most useful, and implementing and testing of the inference algorithm in a PPL.
Titkosítas:
Hozzáférés nincs korlátozva