Neural Networks Overview CS89.11/189.2 - Spring 2020 Our Neurons - - PowerPoint PPT Presentation

neural networks overview
SMART_READER_LITE
LIVE PREVIEW

Neural Networks Overview CS89.11/189.2 - Spring 2020 Our Neurons - - PowerPoint PPT Presentation

Neural Networks Overview CS89.11/189.2 - Spring 2020 Our Neurons Our Neurons Dendrites Our Neurons Dendrites Axon Our Neurons Dendrites Axon terminals Axon Our Neurons Dendrites I 0 I 1 Axon terminals I 2 I 3 Axon I 4 I 5 I m Our


slide-1
SLIDE 1

Neural Networks Overview

CS89.11/189.2 - Spring 2020

slide-2
SLIDE 2

Our Neurons

slide-3
SLIDE 3

Dendrites

Our Neurons

slide-4
SLIDE 4

Axon

Our Neurons

Dendrites

slide-5
SLIDE 5

Axon

Our Neurons

Dendrites Axon terminals

slide-6
SLIDE 6

Axon

Our Neurons

I1 I2 I0 I3 I4 I5 … Im Dendrites Axon terminals

slide-7
SLIDE 7

Axon

Our Neurons

I1 I2 I0 I3 I4 I5 … Im Nucleus decides what to do based on
 the Im and its activation function

Dendrites Axon terminals

slide-8
SLIDE 8

Dendrites Axon

Our Neurons

I1 I2 I0 I3 I4 I5 … Im

Axon terminals

slide-9
SLIDE 9

Axon Axon terminals

Our Neurons

I1 I2 I0 I3 I4 I5 … Im O0 O1 O2 O3 … On

Dendrites

slide-10
SLIDE 10

Artificial Neuron

I0 I1 I2 Neuron Activation Output

slide-11
SLIDE 11

Artificial Neuron

I0 I1 I2 Neuron

Activation Function

Output

i

(Ii × Wi) ϕ(∑

i

(Ii × Wi))

W1 W2 W0

slide-12
SLIDE 12

Artificial Neural Network

slide-13
SLIDE 13

Artificial Neural Network Layers

Input Layer Hidden Layer Output Layer

slide-14
SLIDE 14

Simplified Neural Network

Input Layer Hidden Layer Output Layer

Σ|ϕ Σ|ϕ Σ|ϕ

slide-15
SLIDE 15

Each layer is fully connected to the next

Σ|ϕ Σ|ϕ Σ|ϕ

slide-16
SLIDE 16

A simple pseudo-neural network example

(adapted from Andrew Ng)

Size # Bedrooms Zipcode Average Income Expected Price SchoolQuality Commute FamilyFit

slide-17
SLIDE 17

All neurons in a level have the same

ϕ

ϕ0 ϕ1

Σ|ϕ0 Σ|ϕ0 Σ|ϕ1

slide-18
SLIDE 18

Bias neurons can shift at a level

ϕ

B0 B1

ϕ0 ϕ1

Σ|ϕ0 Σ|ϕ0 Σ|ϕ1

slide-19
SLIDE 19

Neural network examples

B I0 I1 O 1 1

  • 1.5

AND

B I0 I1 O 1 1

  • 0.5

OR

B I0 O

  • 1

0.5

NOT

from AI for Humans, Vol 3., J. Heaton

slide-20
SLIDE 20

Neural network example: XOR

H00

1 1

  • 1.5

adapted from AI for Humans, Vol 3., J. Heaton

1 1

  • 0.5

H10

O 0.5

  • 1

1

  • 1.5

B2 B0 I0 I1

H01

B1 1

I0 I1 H00 H’00 H01 H’01 H10 H’10 O O’

  • 1.5
  • 0.5

0.5 1

  • 0.5

1

  • 0.5

0.5 1 0.5 1 0.5 1 1

  • 0.5

0.5 1 0.5 1 0.5 1 1 1 0.5 1 1.5 1

  • 0.5
  • 0.5

1

ϕ