FU Jun

dxy = [3,4] in your vec it is normalized as [0.6,0,8] for direction, and it loses information of distance. So why not regress directly to [3,4] for both direction and...

can we use an pre-trained embedding instead of training from scratch?

2

Hello, at the very beginning of input, it's a embedding for word2vec from scratch. Can we use an pre-trained embedding instead of training it from scratch? And for the decoding,...

Hello, Pannous, You do great work!!! Really great!!! I'm new comer in speech recognition. I noticed that you cite in the project some papers of 2012 and 2014. Now it...

why recalculate pi and v?

1

Hello, in game_ac_network.py, def prepare_loss(self, entropy_beta), you have: # temporary difference (R-V) (input for policy) self.td = tf.placeholder("float", [None]) value_loss = 0.5 * tf.nn.l2_loss(self.r - self.v) But td == self.r-self.v,...

why batched_weighted_sum?

2

Hello, why use batched_weighted_sum instead of direct dot on all items?

Doesn't work for continuous_mountain_car

6

Hello, the algo doesn't work for continuous_mountain_car, because it's reward is -pow(action[0],2)*0.1. What means, the car's initial state is a local max reward, all the exploration will decrease the reward...

Hello everyone, I put one total week on making this project work. I prove that this project doesn't work at all.

your 15 homegrown noises link zip2 is not valid

1

can not download zip2

Skip connections?

2

Hello, it's great work showing how to use conv_transpose. Do you think the skip connections between encoder layer and decoder layer help? Some papers use it others not (your case)....

resume doesn't load

1

Hello, it's great work! Really great. However, the resume doen't work. Besides, in training, the optimizer is created on model's cpu, however, the model is later put on gpu by...