Tweet
Tweet

Slide 1

Slide 1 text

ML: regularization and neural networks March 2017

Slide 2

Slide 2 text

Recap

Slide 3

Slide 3 text

Recap: linear regression

Slide 4

Slide 4 text

Recap: polynomial regression

Slide 5

Slide 5 text

Recap: gradient descent

Slide 6

Slide 6 text

Regularization

Slide 7

Slide 7 text

Normalization Goal : have all features be equivalent (in size) Rescaling : Standardization :

Slide 8

Slide 8 text

Regularization : goal

Slide 9

Slide 9 text

Regularization Before : Now : Forces values to stay low

Slide 10

Slide 10 text

Neural networks

Slide 11

Slide 11 text

Demo: Tensorflow Playground

Slide 12

Slide 12 text

Biological neuron

Slide 13

Slide 13 text

Artificial neuron: Perceptron

Slide 14

Slide 14 text

Teaching a neuron

Slide 15

Slide 15 text

Artifical neuron : activation function

Slide 16

Slide 16 text

Most common activation functions Identity Sigmoid Tanh ReLU

Slide 17

Slide 17 text

Teaching a neuron

Slide 18

Slide 18 text

Demo: simple network, but the hard way (92%)

Slide 19

Slide 19 text

Demo: simple network, but the hard way (92%)

Slide 20

Slide 20 text

Artificial neuron -> Artificial neural network

Slide 21

Slide 21 text

Artificial neuron -> Artificial neural network

Slide 22

Slide 22 text

Teaching a neural net

Slide 23

Slide 23 text

Demo: Tensorflow Playground

Slide 24

Slide 24 text

Demo: layered network, still the hard way (97%)

Slide 25

Slide 25 text

Questions? March 2017