Teach JS About Aesthetics with Machine Learning - Speaker Deck

Tweet

Tweet

Slide 1

Slide 1 text

Aesthetics + Learning Aesthetics + Learning So long Fine Art.

Slide 2

Slide 2 text

Your trainer. with uncalibrated learning rate

Slide 3

Slide 3 text

Jonathan Big Nerd Ranch – Rails & Frontend

Slide 4

Slide 4 text

@nybblr Pssssst, follow me on Twitter

Slide 5

Slide 5 text

$$$ Google haz them.

Slide 6

Slide 6 text

People. they’re irrational and unpredictable.

Slide 7

Slide 7 text

No content

Slide 8

Slide 8 text

No content

Slide 9

Slide 9 text

No content

Slide 10

Slide 10 text

Decisions. Can’t we automate them?

Slide 11

Slide 11 text

f1 f2 f3 l1 l2 l3

Slide 12

Slide 12 text

No content

Slide 13

Slide 13 text

Instance Everything sounds so wonderful!

Slide 14

Slide 14 text

Features If Ethel orders Lamb Chops?

Slide 15

Slide 15 text

Label Pork chops tonight.

Slide 16

Slide 16 text

Perfect fits But no prediction power.

Slide 17

Slide 17 text

InfoGain Fred has a knack for reducing Lucy’s entropy.

Slide 18

Slide 18 text

Train & Test Take Lucy out for dinner 100x.

Slide 19

Slide 19 text

More $$$ Y Combinator is chatting with your CFO.

Slide 20

Slide 20 text

p1 p2 p3 p4 p5 p6 p7 p8 p9

Slide 21

Slide 21 text

Perceptron p1 f1 f2 f3 out y = f(w1f1 + w2f2 + w3f3)

Slide 22

Slide 22 text

-1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0.5 1 1.5 Activation f(x) = tanh(x)

Slide 23

Slide 23 text

Free Lunch Sorry Lucy, there isn’t one.

Slide 24

Slide 24 text

Aesthetics.js Machines === Artists

Slide 25

Slide 25 text

All the pixels! So long, investors.

Slide 26

Slide 26 text

Composition You really don’t want rabbit ears on center.

Slide 27

Slide 27 text

https://www.ﬂickr.com/ photos/vanheerdenpieter

Slide 28

Slide 28 text

Loneliness if (things.length > 3) blowUp()

Slide 29

Slide 29 text

https://www.ﬂickr.com/photos/wodkawarrior

Slide 30

Slide 30 text

Simplicity You won’t get that low DOF on your iPhone.

Slide 31

Slide 31 text

https://www.ﬂickr.com/photos/wodkawarrior

Slide 32

Slide 32 text

Contrast Vintage faded is so old school.

Slide 33

Slide 33 text

https://www.ﬂickr.com/photos/wodkawarrior

Slide 34

Slide 34 text

Color Theory Red and maroon, really?!

Slide 35

Slide 35 text

https://www.ﬂickr.com/photos/wodkawarrior

Slide 36

Slide 36 text

Wavelets ...to help recruit another investor.

Slide 37

Slide 37 text

Let’s do it! $$$

Slide 38

Slide 38 text

ACQUINE http://wang.ist.psu.edu/IMAGE/

Slide 39

Slide 39 text

No content

Slide 40

Slide 40 text

Gotchas. Minor details like compute time...

Slide 41

Slide 41 text

Brain.js https://github.com/harthur/brain

Slide 42

Slide 42 text

var brain = require('brain'); var net = new brain.NeuralNetwork(); net.train([{input: { r: 0.03, g: 0.7, b: 0.5 }, output: { black: 1 }}, {input: { r: 0.16, g: 0.09, b: 0.2 }, output: { white: 1 }}, {input: { r: 0.5, g: 0.5, b: 1.0 }, output: { white: 1 }}]); var output = net.run({ r: 1, g: 0.4, b: 0 }); // => { white: 0.99, black: 0.002 }

Slide 43

Slide 43 text

ConvNetJS https://github.com/karpathy/convnetjs

Slide 44

Slide 44 text

var layers = [{ type: 'input', out_sx: 1, out_sy: 1, out_depth: 2 }, { type: 'fc', num_neurons: 20, activation: 'relu' }, { type: 'fc', num_neurons: 40, activation: 'relu', drop_prob: 0.5 }, { type: 'softmax', num_classes: 4 }]; var convnetjs = require('convnetjs'); var net = new convnetjs.Net(); net.makeLayers(layers); var input = new convnetjs.Vol(1,1,2); // feature vector var class_probabilities = net.forward(input);

Slide 45

Slide 45 text

Pep talk Machine learning can do anything.

Slide 46

Slide 46 text

No content

Slide 47

Slide 47 text

@nybblr Continue the conversation?