Please enable JavaScript.
Coggle requires JavaScript to display documents.
Transfer Learning in Brains and Brain Inspired Machines (1)Introduciton…
Transfer Learning in Brains and Brain Inspired Machines
1)Introduciton
Introductory statement for a general audience
examples
learn to run after learning how to walk
use of prior knowledge to grasp an unfamiliar object
learn to play cello after mastering violin
learning does not always start from the scratch
transfer learning in the human brain
learning to learn
motor/skill learning
transfer learning in machines
categorisation of transfer learning in ML
state of the art ML methods for TL
what is a task? what is a domain?
limitation in transfer learning machines
filling the gap
transfer learning in brain inspired machines
introduction to BCI
Requirement of TL in BCI
This paper presents
hypothesis and null hypothesis
approach for testing the hypothesis
casestudies
TL Type 1: domain adaptation (same task different domains)
TL Type 2: same domain different tasks (multi-task learning)
hypothesis
Brain Inspired SNN based TL perform better than CNN based TL
null hypothesis
Both algorithms perform equally well and observed differences are merely random
hypothesis testing
two classifiers
Wilcoxon test
t-test
multiple classifiers
ANOVA
Friedman test
main research problem addressed by the paper
introduction to the Brain Inspired TL model using SNN
structure of the paper
3)Transfer Learning in Brain Inspired Machines
Rationale
Brain inspired machines
General introduction to methodology -integrating TL in BI Machines
challenges
how
6) Discussion
casestudy1
statistical validation
casestudy2
statistical validation
hypothesis test
limitations
5) Results
casestudy 1
casestudy 2
7)Conclusion and Future Work
2)Transfer Learning in the Brain
skill learning and motor (re)learning
different types of transfer
learning by observation
learning using previous experience
?
neural mechanisms underlying transfer learning in the brain
neuro-imaging studies
synaptic plasticity
role of neurotransmitters
?
transfer learning hypothesis proposed in previous studies (when+what+where+how transfer happen)
4)Methodology: Proposed Brain Inspired Transfer Learning Model
introduction to methodology
NeuCube SNN framework
knowledge extraction and representation
building the knowledge base for TL
Casestudies
case1
same task different domains
description of dataset and protocol
method
transductive transfer learning
sample selection bias
SNN based approach?
co-variate shift
SNN-based approach?
domain adaptation
SNN-based approach?
Finding model similarity
structural similarity
1 more item...
functional similarity
1 more item...
case2
different tasks same domains
description of dataset and protocol
method - inductive transfer learning
utilize inductive bias
SNN-based approach?
initial connections
Data from task 1
1 more item...
Data from task 2
2 more items...