Jason Yao

Could you provide how you replace `tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits())` without logits? The loss of the original WGAN should not include the sigmoid or softmax functions, it must be linear output, so the...

[Here](https://github.com/hwalsuklee/tensorflow-generative-model-collections/blob/master/WGAN_GP.py) is the WGAN-GP we referenced. We are the WGAN-GP with condition label, but the original WGAN-GP is pure GAN. You are right, you can say that we just implemented...

Cool, thanks for your experiments. Besides the WGAN-GP, you should try spectral normalization for the discriminator, it works like a magic.

You can take a look at the conclusions of this [paper](https://arxiv.org/pdf/1807.04720.pdf) by Google. They said `Our fair and thorough empirical evaluation suggests that when the computational budget is limited one...

The video caption generation is one of the homework in the MLDS (Machine Learning and having it deep and structured) course at NTU. Here is the [link](http://speech.ee.ntu.edu.tw/~tlkagk/courses_MLDS18.html) for this course....

We have updated the README.md for siting the course website at [this commit](https://github.com/JasonYao81000/MLDS2018SPRING/commit/2c5b4791e8255cd27a27726ac9429089f3f7d439). Thanks.

The video features were extracted frame by frame using the pre-trained VGG19 on ImageNet Large Scale Visual Recognition Challenge (ILSVRC). We don't have the code for extracting the video features,...

An episode ends when one of the players reach 21 points in pong. Our reward is defined as: Reward= Nwin - Nlose Nwin is the number of wining rounds in...

I modified the *agent_pg.py* for time recording, and the following is the log of the first 100 training episodes for PPO on the PC with i7-4790 and GTX 1080. The...

Sure, but be aware of the condition (label) for each sample image.