csong27 / property-inference-collaborative-ml Goto Github PK

I cannot run code using newest Theano, it seems incompatible with lasagne in python2

Do I need download this .npz file?

I have my log file attached. Thanks!

Hi, I have some problems with the file ''lfw_attributes.txt''. Where should i download the file? Does the experiment use text dataset of lfw instead of image dataset？I'm a little confused about that. Can you explain the reason? Many Thx!!!

Hi, I would like to know the meaning of test accuracy and p-test accuracy in the following code. Is it the accuracy of the attack or the accuracy of the training task?

for batch in iterate_minibatches(X_test, y_test, 500, shuffle=False):
                inputs, targets = batch
                err, acc = val_fn(inputs, targets)
                val_err += err
                val_acc += acc
                val_batches += 1

            for batch in iterate_minibatches(X_test, p_test, 500, shuffle=False):
                inputs, targets = batch
                err, acc = pval_fn(inputs, targets)
                val_perr += err
                val_pacc += acc
                val_pbatches += 1

            sys.stderr.write("Iteration {} of {} took {:.3f}s\n".format(it + 1, num_iteration,
                                                                        time.time() - start_time))
            sys.stderr.write("  test accuracy:\t\t{:.2f} %\n".format(val_acc / val_batches / 500 * 100))
            sys.stderr.write("  p-test accuracy:\t\t{:.2f} %\n".format(val_pacc / val_pbatches / 500 * 100))
            start_time = time.time()

I am getting this error when I run "distributed_sgd.py":
TypeError: train_multi_task_ps() got an unexpected keyword argument 'balance'

Upon inspecting the code, I realized that on line 117, the call to function "train_multi_task_ps()" includes an argument "balance = balance". But there is no such parameter as "balance" in the definition of "train_multi_task_ps()" on line 340. Can I take out the "balance = balance" argument without any consequences?

Hi, I have some problems with the file ''lfw_attributes.txt''. Where should i download the file? Does the experiment use text dataset of lfw instead of image dataset？I'm a little confused about that. Can you explain the reason? Many Thx!!!

Hi, I try to run the code but it seems load_lfw_with_attrs() has some problems with imgs[common_indices]. I think it might cause by wrong order to run the code? Could you provide the correct order to run the python files in this project?