Closed StudentAlessandro closed 8 months ago
Hey @StudentAlessandro
the epsilon
value in the Epsilon
rule adds a stabilizer constant to the denominator, i.e.
$$R_i = x_i \sumj \frac{w{ji}}{\sum{i'} x{i'} w_{ji'} + b_j + \varepsilon} R_j .$$
Since the denominator normalizes the attainable relevance to $1$, the increased $\varepsilon$ acts as lost relevance (as does the bias $b$), i.e., the sum over all relevances at that layer will be smaller than at the previous layer $\sum_i R_i < \sum_j R_j$. Except for numerical stability, this will not influence your resulting heatmap in usual feed-forward-style networks without parallel layers as you likely normalize the heatmap to its full range before plotting.
Hey @chr5tphr , thank you for the answer, now it's clear.
Hi to everyone I wanted to describe a problem that I encounter when I use the code of the tutorial.
create a composite
choose a target class for the attribution (label 437 is lighthouse)
create the attributor, specifying model and composite
print(f'Prediction: {output.argmax(1)[0].item()}')