contrastive-predictive-coding icon indicating copy to clipboard operation
contrastive-predictive-coding copied to clipboard

Keras implementation of Representation Learning with Contrastive Predictive Coding

Results 16 contrastive-predictive-coding issues
Sort by recently updated
recently updated
newest added

I try many times to repeat the result, but my binary_accuracy always 50%, it's horrible. then i find that an issue mention that the positive_sample is the half of batch_size....

Has anyone found a way (or a set of fixed random seed) to reproduce the results? I tried to fix every random seed, but it seems that there still exists...

Hi @davidtellez, This is a great project to get started with CPC; however, while running this project in my local environment, I found that while most of the code is...

hello,i have a question about predict_terms and terms.Could you please tell me what these mean?

It would be nice to run a MLP on the encoder representation to compare the representation learned by the unsupervised encoder in comparison to the full CPC model representation.

Hi David, Its a great repo. I have doubt regarding the final 90% accuracy that you received on the MNSIT dataset in 10 epochs. Isn't that necessarily because of the...

As discussed in the original paper, the training relies on a large number of negative pairs to tighten the lower bound of the Mutual Information, which corresponds to log(N). However,...

Hello, thank you very much for the code you provided. Why did you make a toy data set to do CPC training, and not just use the original minst data...

Hi, nice code, but I'm confused about the training code of encoder. I think the training of encoder via CPC is unsupervised, however, you used some label information (made by...

Hi @davidtellez, In the readme you report the accuracy of CPC is 99%. But I got ~50%. Here my log: ``` poch 1/10 390/390 [==============================] - 189s 484ms/step - loss:...