Layerwise Relevance Propagation with Deep Taylor Series in TensorFlow
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.ipynb_checkpoints added code and demo Dec 7, 2016
LRP_Example.ipynb added code and demo Dec 7, 2016
README.md Update README.md Jan 22, 2017
lrp.py added code and demo Dec 7, 2016
lrp.pyc added code and demo Dec 7, 2016

README.md

Tensorflow_Deep_Taylor_LRP

Layerwise Relevance Propagation with Deep Taylor Series in TensorFlow.

You can use LRP to visualize the relative feature importances of the input to a neural network.

How to Use

Step 1: Construct your tensorflow graph

Step 2: Make sure your prediction layer (output layer) is named "absolute_output"

Step 3: Make sure your input layer (shaped as [num_batches, height, width, num_channels]) is named "absolute_input"

Step 4: relevance_heatmap = lrp.lrp(prediction*label, lowest_value, highest_value)