CS6501: Deep Learning for Visual Recognition Recurrent Neural - - PowerPoint PPT Presentation

▶
cs6501 deep learning for visual recognition
SMART_READER_LITE
LIVE PREVIEW

CS6501: Deep Learning for Visual Recognition Recurrent Neural - - PowerPoint PPT Presentation

Feb 05, 2024 565 likes •794 views

CS6501: Deep Learning for Visual Recognition Recurrent Neural Networks (RNNs) Todays Class Recurrent Neural Network Cell Recurrent Neural Networks (RNNs) Bi-Directional Recurrent Neural Networks (Bi-RNNs) Multiple-layer /

slide-1
SLIDE 1

CS6501: Deep Learning for Visual Recognition

Recurrent Neural Networks (RNNs)

slide-2
SLIDE 2
  • Recurrent Neural Network Cell
  • Recurrent Neural Networks (RNNs)
  • Bi-Directional Recurrent Neural Networks (Bi-RNNs)
  • Multiple-layer / Stacked / Deep Bi-Direction Recurrent Neural

Networks

  • LSTMs and GRUs.
  • Applications in Vision: Caption Generation.

Today’s Class

slide-3
SLIDE 3

Yann LeCun CNNs ACM Turing Award 2019 Yoshua Bengio GANs Geoff Hinton Backpropagation

slide-4
SLIDE 4

Recurrent Neural Network Cell

!"

#$$

ℎ& ℎ"

slide-5
SLIDE 5

Recurrent Neural Network Cell

!"

#$$

ℎ& ℎ" ℎ" = tanh (.

//ℎ& + . /1!")

slide-6
SLIDE 6

Recurrent Neural Network Cell

!"

#$$

ℎ& ℎ" ℎ" = tanh (.

//ℎ& + . /1!")

ℎ" 3" 3" = softmax (.

/9ℎ")

slide-7
SLIDE 7

Recurrent Neural Network Cell

!"

#$$

ℎ& ℎ" ℎ" '"

slide-8
SLIDE 8

Recurrent Neural Network Cell

!""

#$ = [0 0 1 0 0] ℎ, = [0 0 0 0 0 0 0 ]

  • $ = [0.1, 0.05, 0.05, 0.1, 0.7]

ℎ$ = [0.1 0.2 0 − 0.3 − 0.1 ] ℎ$ = [0.1 0.2 0 − 0.3 − 0.1 ]

a b c d e

e (0.7)

c

slide-9
SLIDE 9

Recurrent Neural Network Cell

!"

#$$

ℎ& ℎ" ℎ" '"

slide-10
SLIDE 10

Recurrent Neural Network Cell

!"

#$$

ℎ& ℎ" ℎ"

slide-11
SLIDE 11

(Unrolled) Recurrent Neural Network

!"

#$$

ℎ& ℎ" ℎ" !'

#$$

ℎ' ℎ' !(

#$$

ℎ( ℎ( c a t a t <<space>> )" )' )(

slide-12
SLIDE 12

(Unrolled) Recurrent Neural Network

!"

#$$

ℎ& ℎ" ℎ" !'

#$$

ℎ' ℎ' !(

#$$

ℎ( ℎ( the cat likes cat likes eating )" )' )(

slide-13
SLIDE 13

(Unrolled) Recurrent Neural Network

!"

#$$

ℎ& ℎ" !'

#$$

ℎ' !(

#$$

ℎ( ℎ( the cat likes positive / negative sentiment rating )

slide-14
SLIDE 14

Bidirectional Recurrent Neural Network

!"

#$%%

ℎ' ℎ" ℎ" !(

B$%%

ℎ( ℎ( !*

#$%%

ℎ* ℎ* the cat wants gato quiere comer +" +( +*

slide-15
SLIDE 15

Stacked Recurrent Neural Network

!"

#$$

ℎ &" !'

#$$

!(

#$$

c a t )" )' )(

#$$

ℎ* ℎ" ℎ"

#$$

ℎ' ℎ'

#$$

ℎ( ℎ( ℎ &' ℎ &( ℎ &* ℎ &" ℎ &' ℎ &(

slide-16
SLIDE 16

Stacked Bidirectional Recurrent Neural Network

!"

#$$

ℎ &" !'

#$$

!(

#$$

c a t )" )' )(

#$$

ℎ* ℎ" ℎ"

#$$

ℎ' ℎ'

#$$

ℎ( ℎ( ℎ &' ℎ &( ℎ &* ℎ &" ℎ &' ℎ &(

slide-17
SLIDE 17

RNN in Pytorch

slide-18
SLIDE 18

LSTM Cell (Long Short-Term Memory)

!"

#$%&

ℎ( ℎ" )( )"

slide-19
SLIDE 19

LSTM in Pytorch

slide-20
SLIDE 20

GRU in Pytorch

slide-21
SLIDE 21

Questions?

21