Commit 15aa44f9 authored by Ying-Qiu Zheng's avatar Ying-Qiu Zheng
Browse files

udpate notes

parent 190b3259
......@@ -2,7 +2,18 @@
![](figs/all_vs_best100_reg.png)
## Summary
### comparison of different basis, ICA, LE, and PCA - how to find better basis? - more advanced methods may come in...(factorise resting data into modes..)
### 1) What is the set of bases that can better represent task-activation in individuals? - how to find a better set of bases? - more advanced methods may come in...(factorise resting data into modes..)
* for now we focus on the comparison of simple methods, i.e., ICA, PCA, Laplacian eigenmaps (with number of components predetermined)
* 100 components
![](figs/basis_comparison_boxplot.png)
* 50 components
* automatically find the number of components? Bayesian methods?
* For now the basis only used information from resting state data of the subject and other subjects, maybe try to incorprate information from other subjects' task activation information?
### 2) using residual basis of the self subject and the coefficients of other subjects to make predictions
* Using residual bases to make prediction differentiates self from others more than using original basis
<img src="figs/improv.png" width="49%"> <img src="figs/improv_reg.png" width="49%">
* How to further improve the performance?
### 3) Amplitude prediction
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment