-
- iNNvestigate explainability software
- [x] read [iNNvestigate paper](https://arxiv.org/pdf/1808.04260.pdf)
- [x] install [iNNvestigate](https://github.com/albermax/innvestigate)
- [x] stud…
-
https://nbviewer.jupyter.org/github/UntangleAI/example/blob/master/stylized_imagenet_vis_check_alexnet.ipynb
If you still want to check it out.
Would appreciate insights as to why stylised image…
-
Hi AlberMax,
I have problems using the heatmap methods in iNNvestigate for neuroimaging models/data. So can you please let me know, for now which methods in iNNvestigate support 3DCNN architectures…
-
Hi, I am very appriciate to your code. Thank you so much.
I have two questions about PatternNet:
1. what's the quantity that is backproparated? Is it 'a'?
2. What's the backpropagation algorithm?…
-
Hi,
Thank you for open sourcing this package. It must be a lot of work. I have some questions.
I am trying to use PatternNet and PatternAttribution with any predefined network, i.e., if I defi…
-
Hi,
I am wondering which of the methods proposed work with networks that use normalization layers or whether there are any other constraints to the model architecture. From the PatternNet paper it…
-
* get to work for all application networks.
* harden code:
* warn when used with not specified layers networks.
* check if some kernel layers have no pattern (needed?).
-
Adding Gradient*Input to the list of methods in `examples/mnist_select_neuron.py` yields an `AttributeError`:
```
Output Neuron 0: Input (0.0641s) Gradient (0.3563s) Deconvnet (0.5039s) Guided Bac…
-
Hi,
thanks for sharing this code!
Could you add some code for computing the patterns for PatternNet etc.? That would be very helpful.
Thanks,
Robin
-
I have the newest Keras 2.1.6 and all_methods.py crashes for PatternNet (`TypeError: softmax() got an unexpected keyword argument 'axis'` etc.). However, it works with Keras 2.1.5.
Tensorflow was o…