Carlo David

add papers

1

added latest papers on differential privacy with deep learning

There's a typo on the additional resources of Lesson 4 concept 5, which says, "For a beginner-friendly explanation on `holomorphic` encryption, see What is Homomorphic Encryption?". It should be homomorphic...

I tried to train text classifier using LSTM and the accuracy is always 0.51 and below

low accuracy when using DPAdamGaussianOptimizer

1

When I train on MNIST data using DPAdamGaussianOptimizer, i got very low accuracy. But it works fine when using DP-SGD. Code: ``` optimizer = DPAdamGaussianOptimizer( l2_norm_clip=1.5, noise_multiplier=1.3, num_microbatches=250, learning_rate=0.25) loss...

what is the optimal RDP order?

4

when i use compute_dp_sgd_privacy, it shows this result. I know the epsilon is the strength of the privacy guarantee, but what does optimal RDP order means? ``` compute_dp_sgd_privacy.compute_dp_sgd_privacy(n=len(training_sequences), batch_size=250, noise_multiplier=1.3,...