1 / 15

Learning, development and plasticity Josh Tenenbaum MIT Department of Brain and Cognitive Sciences

Learning, development and plasticity Josh Tenenbaum MIT Department of Brain and Cognitive Sciences Computer Science and AI Lab (CSAIL). Can we develop an integrated theory of learning, in computational, behavioral and neural terms?. The “standard model”. Supervised. Unsupervised.

tolla
Télécharger la présentation

Learning, development and plasticity Josh Tenenbaum MIT Department of Brain and Cognitive Sciences

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Learning, development and plasticity Josh Tenenbaum MIT Department of Brain and Cognitive Sciences Computer Science and AI Lab (CSAIL)

  2. Can we develop an integrated theory of learning, in computational, behavioral and neural terms? • The “standard model” Supervised Unsupervised

  3. The really hard problems

  4. The two cultures “Nature” Innate structured representations Grounded in cognition versus “Nurture” Statistical learning, plasticity Grounded in the brain

  5. Recent causes for optimism • New models from machine learning, AI • Structured statistical models Probabilities defined over structured representations: graphs, causal networks, grammars, predicate logic. • Multilevel (hierarchical) statistical models Inference at multiple levels of abstraction and multiple timescales. • Flexible statistical models Hypothesis spaces grow as new data are encountered. • New technologies • “Supercomputers” on the desktop, grid computing • Life-size datasets for modeling cognitive development • Mainstream functional MRI

  6. P(grammar | UG) P(phrase structure | grammar) P(utterance | phrase structure) P(speech | utterance) “Universal Grammar” Hierarchical phrase structure grammars (e.g., CFG, HPSG, TAG) Grammar Left IFG Phrase structure Utterance Speech signal

  7. t=0 Dest=2 2 2 Dest=2 1 1 Dest=1 t=1 Dest=1 t=2 Probabilistic scene parsing BLOG (Bayesian Logic)

  8. Probabilistic scene parsing

  9. Learning domain structures F: form Tree with objects at leaf nodes pine P(structure | form) tree plant palm S: structure living thing mouse animal mammal P(data | structure) gorilla D: data

  10. Learning domain structures

  11. Learning causal theories Abstract Principles Structure Data

  12. Learning regulatory modules in genetics Learning the laws of magnetism Causal learning with complementary “cortical” and “hippocampal” networks

  13. Learning to act Exploration vs. Exploitation Concepts learned: Clear, Inhand Topstack Above Height …

  14. Constraints Goals Rational planning (PO)MDP Actions Understanding actions:goal inference Human judgments Model Predictions

  15. The grand challenge Cognitive science of human learning Design of artificial learning systems Brain structures and mechanisms that support learning

More Related