0%

Tutorial:Decoding to Understand Neural

1. Motication

  • have some idea about how brain works
  • run a experiment and make neural recordings
  • get a buch of data

convert data into answers

2. Decoding?

Alternative name: readout MVPA

predict stimulus/behavior from neural activity

Classifier:

  1. see some image, record the eeg
  2. train pattern classifier
  3. predict

this a analog to the real brain

image-20200406185816001

3. Neural Content

Start with a simply experiment

Confusion Matrices, predict pretty well

different classifiers? both good

  • MCC
  • EDist(good)
  • SVM

3.1. Abstract representations

Different angles/language/color, we can still tell

Example: position invariance experiment

  • train upper, better accuracy upper, but not bad at middle and lower
  • different pose ML/MF/AL/AM

4. Data analsis

neural content: what information

coding: what feature does brain process