Deep Learning Flashcards by Arcady Dushatsky | Brainscape

Q

A

How well did you know this?

1

Not at all

2

3

4

5

Perfectly

Q

what is stochastic in SGD?

A

batches, initial weights

How well did you know this?

1

Not at all

2

3

4

5

Perfectly

Q

what is momentum doing?

A

makes sgd more stable, m=0.99m+0.01gradient

How well did you know this?

1

Not at all

2

3

4

5

Perfectly

Q

why is relu so popular?

A

fast to compute, no vanishing gradients

How well did you know this?

1

Not at all

2

3

4

5

Perfectly