search

mbhenaff / spectral-lib

45 stars 18 forks source link

testing #6

Open gravity1989 opened 8 years ago

gravity1989 commented 8 years ago

So I managed to train using the cifar dataset.

How do i test the model using the test dataset?

mbhenaff commented 8 years ago

Hi Rahul,

The test performance is computed line 69 of test/train.lua. The computePerf function takes in a split ('train' or 'test') and computes performance of the model on that split.

-Mikael

On Tue, Mar 29, 2016 at 1:52 AM, Rahul V notifications@github.com wrote:

So I managed to train using the cifar dataset.

How do i test the model using the test dataest?

— You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub https://github.com/mbhenaff/spectral-lib/issues/6

gravity1989 commented 8 years ago

Thanks Mikael.

On Wed, Mar 30, 2016 at 7:10 AM, mbhenaff notifications@github.com wrote:

Hi Rahul,

The test performance is computed line 69 of test/train.lua. The computePerf function takes in a split ('train' or 'test') and computes performance of the model on that split.

-Mikael

On Tue, Mar 29, 2016 at 1:52 AM, Rahul V notifications@github.com wrote:

So I managed to train using the cifar dataset.

How do i test the model using the test dataest?

— You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub https://github.com/mbhenaff/spectral-lib/issues/6

— You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub https://github.com/mbhenaff/spectral-lib/issues/6#issuecomment-203185327

Regards, R.V.

wangg12 commented 8 years ago

@gravity1989 I have a question. Is the Laplacian matrix obtained from all the training data and fixed during training? If a single sample is a graph structured data(say represented as a weighted matrix), how should I do?

gravity1989 commented 8 years ago

I am not sure if its fixed or otherwise. I have just about managed to runt he code.

The .mat file is loaded and modifications could be happening which is invisible to me.

On Fri, Apr 1, 2016 at 1:02 PM, Wang Gu notifications@github.com wrote:

@gravity1989 https://github.com/gravity1989 I have a question. Is the Laplacian matrix obtained from all the training data and fixed during training? If a single sample is a graph structured data(say represented as a weighted matrix), how should I do?

— You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub https://github.com/mbhenaff/spectral-lib/issues/6#issuecomment-204290660

Regards, R.V.

mbhenaff commented 8 years ago

Yes, the Laplacian matrix is fixed during training. It is first estimated from the training data, and then its eigenvectors are used to define a Graph Fourier Transform which is a generalization of the regular Fourier transform. Pointwise multiplication in this new "frequency" space corresponds to a convolution on the graph.

On Fri, Apr 1, 2016 at 3:38 AM, Rahul V notifications@github.com wrote:

I am not sure if its fixed or otherwise. I have just about managed to runt he code.

The .mat file is loaded and modifications could be happening which is invisible to me.

On Fri, Apr 1, 2016 at 1:02 PM, Wang Gu notifications@github.com wrote:

@gravity1989 https://github.com/gravity1989 I have a question. Is the Laplacian matrix obtained from all the training data and fixed during training? If a single sample is a graph structured data(say represented as a weighted matrix), how should I do?

— You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub < https://github.com/mbhenaff/spectral-lib/issues/6#issuecomment-204290660>

Regards, R.V.

— You are receiving this because you commented. Reply to this email directly or view it on GitHub https://github.com/mbhenaff/spectral-lib/issues/6#issuecomment-204292418