Kari's user avatar
Kari's user avatar
Kari's user avatar
Kari
  • Member for 6 years, 2 months
  • Last seen more than a month ago
  • UK
22 votes
2 answers
17k views

Sliding window leads to overfitting in LSTM?

14 votes
2 answers
5k views

Prioritized Replay, what does Importance Sampling really do?

14 votes
2 answers
3k views

Why averaging the gradient works in Gradient Descent?

11 votes
2 answers
11k views

RL Advantage function why A = Q-V instead of A=V-Q?

11 votes
2 answers
3k views

How does Implicit Quantile-Regression Network (IQN) differ from QR-DQN?

11 votes
1 answer
177 views

Synthetic Gradients good number of Layers & neurons

11 votes
1 answer
8k views

Q-Learning: Target Network vs Double DQN

8 votes
2 answers
908 views

Synthetic Gradients - what's the practical benefit?

7 votes
2 answers
9k views

Muti-hot encoding vs Label-Encoding

5 votes
1 answer
237 views

Policy Gradients - gradient Log probabilities favor less likely actions?

5 votes
2 answers
229 views

Gradient Checking LSTM - how to get change in Cost across timesteps?

5 votes
2 answers
3k views

Cross Validation how to determine when to Early Stop?

4 votes
2 answers
844 views

Prioritized Experience Replay - why to approximate the Density Function?

4 votes
1 answer
2k views

Why does momentum need learning rate?

4 votes
1 answer
1k views

Policy-based RL method - how do continuous actions look like?

4 votes
1 answer
3k views

Dueling DQN - can't understand its mechanism

4 votes
2 answers
3k views

Auto-Encoder to condense (pre-process) large one-hot input vectors?

4 votes
1 answer
633 views

How does action get selected in a Policy Gradient Method?

4 votes
2 answers
966 views

Hindsight Experience Replay, how to define a partially-known End-Goal

4 votes
1 answer
1k views

selecting a number of neurons specifically for RL

3 votes
1 answer
742 views

Rainbow vs A3C ...too unfair?

3 votes
1 answer
329 views

Dueling DQN what does a' mean?

3 votes
1 answer
2k views

Policy Gradients vs Value function, when implemented via DQN

3 votes
1 answer
231 views

GradientChecking, can I blame float precision?

3 votes
2 answers
273 views

RL: Collecting States (training data) in real-life. Must use fixed timestep?

3 votes
1 answer
91 views

Gradient Checking: MeanSquareError. Why huge epsilon improves discrepancy?

3 votes
2 answers
2k views

Why RMSProp converges faster than Momentum?

3 votes
1 answer
66 views

Grad Checking, verify by average?

3 votes
1 answer
439 views

Neural Turing Machine - how the shifting works?

3 votes
1 answer
254 views

RNN: why Wx + Uh instead of W[x,h]