The Science of Learning Breaking News D December 2010 b 2010 Joan - - PowerPoint PPT Presentation

The Science of Learning Breaking News D b 2010 December 2010

Joan Straumanis

Current SLC Portfolio

2004 Cohort  CELEST: Center for Cognitive d Ed ti l N i 2006 Cohort  SILC: Spatial Intelligence and L i C t (T l U) and Educational Neuroscience (Boston U) http://celest.bu.edu Learning Center (Temple U) http://spatiallearning.org  TDLC: Temporal Dynamics of  LIFE: Learning in Informal and Formal Environments (U of Washington) http://www.life‐ slc.org  TDLC: Temporal Dynamics of Learning Center (UC‐San Diego) http://tdlc.ucsd.edu slc.org  PSLC: Pittsburgh Science of Learning Center (Carnegie‐ Mellon U)  VL2: Visual Language and Learning Center (Gallaudet U) Mellon U) http://www.learnlab.org ( ) http://vl2.gallaudet.edu

Definition of Robust Learning:

• Retained for a long time
• Effective preparation for further learning or

practice practice f l

• Transfers to novel situations

To Increase Learning:

 Enlist brain’s motivational & reward systems; compete effectively with other rewards  Manage sleep to consolidate memory  Multimodal input  Ensure engagement (“Active Learning”)  Ensure engagement ( Active Learning )  Manage timing of practice, reinforcement (Assistance Dilemma)  Provide plenty of social interaction!

What we’ve learned about STEM learning STEM learning…

(a synthesis)

 Expert explanation……….is not as effective as  Peer explanation …which is not as effective as  Self explanation which is not as effective as  Self explanation…..which is not as effective as  Teaching another…even when that other is a Teaching another…even when that other is a computer‐generated avatar

Findings about Social Interaction:

• Babies need it to learn language
• Teaching another is powerful so
• Teaching another is powerful…so
• Team/group learning is effective
• “Mere belief in the social” works!
• But so does teaching a computer avatar!!

Social Dynamics

• f Learning:
• f Learning:

Human, Robot, Vid Video

UW MEG Brain Imaging Center

LIFE reported the first in the world MEG recordings of awake infants engaged in a cognitive task (Imada Kuhl cognitive task (Imada, Kuhl et al., 2006). Magnetic fields generated by neuronal activity in the brain are recorded

• utside the brain by magnetometers.

8

Ribbon-Cutting on May 24, 2010

International collaboration with Helsinki University of Technology and Elekta Oy – all 6 SLCs in Helsinki in Dec., 2009

Science at CELEST

• Focus: Brain mechanisms of learning during planning,

exploring, communicating & remembering

• Approach: Integration of modeling & experimentation
• Cross‐Cutting Research Themes: processing bottlenecks ,

dynamic coding functional connections neural plasticity dynamic coding, functional connections, neural plasticity, neuromorphic engineering

• Examples: Links between gamma oscillations, plasticity &

learning to study phase coding in working memory, illuminating cognitive bottlenecks; biologically inspired microprocessing using memristors using memristors

The DARPA SyNAPSE Project

A National Science Foundation Science of Learning Center Center of Excellence for Learning in Education, Science, and Technology

The DARPA SyNAPSE Project Hardware goals

Systems of Neuromorphic Adaptive Plastic Scalable Electronics 106 “neurons”/cm2 1020 “synapses”/cm2 ~100 milliwatts/cm2 10,000 chips, 1000 watts stuffed in a shoebox

To date: Wrong hardware!

Brain-like computations with CELEST i ki i h Brain like computations with co-located DATA STORAGE, SIGNAL TRANSMISSION, and LEARNING are inefficient in di it l t CELEST is working with Industrial partners HRL Labs and Hewlett-Packard to devise next-generation computing hardware. digital computers

1010 neurons 1014 synapses

next generation computing hardware. Such machines MUST LEARN, because they will not be fully “programmable”.

Blue Gene 1 GW, thousands

• f racks

Brain 20 W 1.3Kg

Phase‐Coding of Objects in Working Memory

What are the fundamental constraints on working memory capacity?

This analysis was directly inspired by CELEST interactions between members of the Hasselmo and Miller labs. Koene, R. & Hasselmo, M. (2007). Cerebral Cortex, 17, 1766–1781.

Spikes associated with the first and second Spikes associated with the first and second items occur at different phases within local field potentials that cycle at approximately 32 Hz i h f l f k in the prefrontal cortex of monkeys performing a working memory task.

11

Siegel, Warden, and Miller (2009) Proc. Nat. Acad. Sci.

TDLC:Time matters for processing…

“say” y

Frequency

“stay”

F

100 ms

Ti ( illi d ) Time (milliseconds) These wave forms are id entica l id entica l except for the artificially inserted gap and a compensating shrinkage of the waveform.

Facial Expression during problem solving. Littlewort, Phan Reilly and Bartlett (3 2) Phan, Reilly, and Bartlett, (3.2)

Facial Expression during problem solving.

Littlewort, Phan, Reilly, and Bartlett, (3.2)

Initiative 6- Translational Research Initiative 6- Translational Research