cnn_graph
cnn_graph copied to clipboard
Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering
The usage file and another paper I found mentions the possibility of using the architecture for node classification. I wondered how the signal classification example differs from node classification with...
When approximate the initial filters kernel gθ(Λ) = diag(θ) with a polynomial ![image](https://user-images.githubusercontent.com/40826585/44521567-1512dc00-a706-11e8-8219-2b85f0f71002.png) why we can‘t simply combine the eigenvector and eigenvalue matrix together like ![image](https://user-images.githubusercontent.com/40826585/44521653-5efbc200-a706-11e8-8920-4dcd9c3e2c5d.png) because ![image](https://user-images.githubusercontent.com/40826585/44521667-6ae78400-a706-11e8-89f8-af9c3b2a941a.png) * The...
In the module [`graph.py`](https://github.com/mdeff/cnn_graph/blob/master/lib/graph.py#L64), you are calculating a variable `sigma2` as follows `sigma2 = np.mean(dist[:, -1])**2`. However, this is "mu squared" (not the variance) of the last column of the...
First off, thanks for developing and sharing this interesting package. I've forked the repo and all test cases work fine. This is probably related to #14, but I've made a...
Hi! First of all thanks for creating such a nice reproducible example for your repo, you made your algorithm way more accessible. I have a question regarding accessing the output...
At the following line, https://github.com/mdeff/cnn_graph/blob/master/lib/coarsening.py#L210, you have the assertion `assert len(indices[0] == M)`, but this is always true, because `indices[0]` is not empty and `indices[0] == M` produces another list...
## Hi there! I've tried for a while to run your code, but still have some flag problems, so here what I get: **_this is what I get if I...
Hi, I'm wondering if there exists or if your are planning to implement any version of your work in Keras. Thank you Andrea
Hello @mdeff , MNIST data are defined on 2D grid. Hence, we build graph on MNIST by supposing that each pixel is a node and the max number edges per...
Hello @mdeff Let me thank you for this notebook showing different graph convolution implementation. https://github.com/mdeff/cnn_graph/blob/master/trials/1_learning_filters.ipynb l'm wondering if you have an optimized implementation of a convolution in a full spectral...