Skip to content

The Visualization of Decision Making -- in Neuroimaging Classifiers

Notifications You must be signed in to change notification settings

jzm0144/neuroVis_old

Repository files navigation

This project uses Python=3.6

Please install all the required Libariers for the project using the following command.

$ pip install -r requirements.txt 

Please run the file main.py with the four arguments: the dataset, the brain disorder, number of paths, the class id

Here is an example run:

$ python main.py ageMatchUnmatch disorder dataExample topPaths classID
$ python main.py ageMatched         PTSD      13         10       0

Part1 Results:

Part1 results are in Results/Part1/PTSD/. Generates specific explanation for each heatmap method and each example.

Part2 Results:

Part2 results are in Results/Part2/PTSD/. In Part2 experiment we calculate an avg of each Heatmap Method over all data examples. It generates a 1 final image (avg of heatmaps of all examples) for each heatmap method.

Part3 Results:

Part3 results are in Results/Part3/PTSD/. It generates an avg of All Heatmaps for a given example. So this kind of explanation takes the opinion of each heatmap method but in a very naive way.

Part4 Results:

Part4 results are in Results/Part4/PTSD/. This experiment calculates an explanation based on the opinion of all heatmaps.
--------- step1: Calc all Heatmaps for the Same Example
--------- step2: Sparsify: Let only top-X paths pass thruough
--------- step3: Calc Binary Intersection of path occurences
--------- step4: Calc Mean and Element-wise multiply with Binary Intersetion

About

The Visualization of Decision Making -- in Neuroimaging Classifiers

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages