Implicit Statistical Machine Learning A Common Mechanism or

  • Slides: 15
Download presentation
Implicit, Statistical, Machine Learning: A Common Mechanism or Many? Paul J. Reber Department of

Implicit, Statistical, Machine Learning: A Common Mechanism or Many? Paul J. Reber Department of Psychology Northwestern University

IMPLICIT LEARNING OF REPEATING SEQUENCES • …

IMPLICIT LEARNING OF REPEATING SEQUENCES • …

REPEATING SEQUENCE LEARNING = TRIGRAM LEARNING • …

REPEATING SEQUENCE LEARNING = TRIGRAM LEARNING • …

FRAGMENT TASK 6 blocks Fragment based training Repeated sequence training 80% S 1 block

FRAGMENT TASK 6 blocks Fragment based training Repeated sequence training 80% S 1 block 540 test trials 33% S

FRAGMENT DATA (reanalysis) • N=374

FRAGMENT DATA (reanalysis) • N=374

ASISL TASK 1 6 blocks 1 block 540 training trials 540 test trials 80%

ASISL TASK 1 6 blocks 1 block 540 training trials 540 test trials 80% S 33% S s 45 reps of alternating sequence/training block 1 r 2 r 3 r 4 r 5 r 6 r d f j k l

ASISL TASK 2 6 blocks 1 block 540 training trials 540 test trials 80%

ASISL TASK 2 6 blocks 1 block 540 training trials 540 test trials 80% S 33% S s 45 reps of alternating sequence/training block 6 transfer trigrams d f j k l

ASISL DATA 1 Training data: this figure will be replaced Sequencespecific learning at test=3.

ASISL DATA 1 Training data: this figure will be replaced Sequencespecific learning at test=3. 5%, t(29)=2. 9, p<0. 01 *

ASISL DATA 2 All of these figures will be replaced

ASISL DATA 2 All of these figures will be replaced

IMPLICIT LEARNING IS INFLEXIBLE • …

IMPLICIT LEARNING IS INFLEXIBLE • …

AB SISL TASK 2 blocks 1 block A B A/B 540 training trials 540

AB SISL TASK 2 blocks 1 block A B A/B 540 training trials 540 test trials 80% S 33% S

AB SISL DATA • N=38 • Sequence 1 SSPA is reliably greater than 0,

AB SISL DATA • N=38 • Sequence 1 SSPA is reliably greater than 0, t(37)=5. 4, p<0. 001 • Sequence 2 SSPA is reliably greater than 0, t(37)=3. 7, p<0. 001 • No significant difference in SSPA between Sequences 1 and 2, t(37)=0. 19, p=0. 8

SUMMARY • …

SUMMARY • …

REBERLAB: Cognitive Neuroscience of Expertise KEVIN SCHMIDT • High-repetition simulationbased training: topographic maps •

REBERLAB: Cognitive Neuroscience of Expertise KEVIN SCHMIDT • High-repetition simulationbased training: topographic maps • Modeling memory system interactions: PINNACLE • Reward, cognitive control and memory • Visually-cued perceptual-motor sequence learning: transfer, BROOKE FEINSTEIN BEN REUVENI flexibility, individual differences MICHAEL COHEN EVAN GRANDOIT • Auditorily perceptual-motor sequence learning PEIGEN SHU KELSEY THOMPSON CAT HAN