en-de
en-es
en-fr
en-sl
en
en-zh
0.25
0.5
0.75
1.25
1.5
1.75
2
Bayesian models of human inductive learning
Published on 2007-06-2227456 Views
In everyday learning and reasoning, people routinely draw successful generalizations from very limited evidence. Even young children can infer the meanings of words, hidden properties of objects, or t
Related categories
Presentation
Bayesian models of human inductive learning Josh Tenenbaum MIT Department of Brain and Cognitive Sciences Computer Science and AI Lab (CSAIL)00:00
Lab members02:48
The probabilistic revolution in AI 03:20
Everyday inductive leaps05:00
Learning concepts from examples06:36
The solution09:22
The solution0110:08
The approach: from statistics to intelligence14:10
Outline16:33
The “shape bias” in word learning (Landau, Smith, Jones 1988)17:12
Is the shape bias learned?18:54
Transfer to real-world vocabulary21:01
Learning about feature variability24:06
Learning about feature variability0124:52
A hierarchical Bayesian model25:25
A hierarchical Bayesian model0126:07
A hierarchical Bayesian model0227:13
A hierarchical Bayesian model0327:52
A hierarchical Bayesian model0428:50
Learning the shape bias29:35
Learning the shape bias0130:23
Extensions31:43
Learning to transfer selectively33:02
Learning to transfer selectively0134:12
Property induction36:29
The computational problem37:59
Hierarchical Bayesian Framework40:45
P(D|S): How the structure constrains the data of experience42:22
P(D|S): How the structure constrains the data of experience0142:53
P(D|S): How the structure constrains the data of experience0244:03
Structure S44:14
slide3444:57
[c.f., Lawrence, 2004; Smola & Kondor 2003]45:10
Cows have property P. Elephants have property P. Horses have property P.46:09
Testing different priors47:29
Learning about spatial properties 49:18
Discovering structural forms50:44
Discovering structural forms0151:04
People can discover structural forms51:24
The ultimate goal52:52
A “universal grammar” for structural forms53:20
slide4755:32
Structural forms from relational data57:23
Lab studies of learning structural forms57:31
Development of structural forms as more data are observed57:36
Beyond “Nativism” versus “Empiricism”01:00:09
Summary01:01:44