Slide 1

Slide 1 text

Convolutional Neural Networks & Galaxies Ryan Keisler, Stanford “Computing the Universe”, UC Berkeley, Jan. 2015

Slide 2

Slide 2 text

David W. Hogg and the Sloan Digital Sky Survey Collaborations.! All images are © 2013 David W. Hogg.

Slide 3

Slide 3 text

> (5 numbers) How to quantify beyond fluxes?

Slide 4

Slide 4 text

…using machines

Slide 5

Slide 5 text

…using people

Slide 6

Slide 6 text

Galaxy Zoo 2: detailed morphological classifications for ! 304,122 galaxies from the Sloan Digital Sky Survey! ! K. Willett et al 2013 • ~300e3 galaxies (mr > 17) • ~16e6 classifications • ~50 classifications / galaxy 100s of 1000s of volunteers!

Slide 7

Slide 7 text

Classification Decision Tree 37 different outcomes

Slide 8

Slide 8 text

0 B B B B B B B B B B @ n1 . . . . . . . . . . . . n37 1 C C C C C C C C C C A 3 x 96 x 96 = 27648 numbers 37 numbers People (even “pro astronomers”) disagree, so there is no single classification per galaxy.
 
 The distribution of votes is the “classification”.

Slide 9

Slide 9 text

0 B B B B B B B B B B @ n1 . . . . . . . . . . . . n37 1 C C C C C C C C C C A 3 x 96 x 96 = 27648 numbers 37 numbers Thanks for the labels, volunteers! ! Now time for some supervised machine learning. some algorithm

Slide 10

Slide 10 text

kaggle! a website for machine learning competitions

Slide 11

Slide 11 text

0 B B B B B B B B B B @ n1 . . . . . . . . . . . . n37 1 C C C C C C C C C C A 3 x 96 x 96 = 27648 numbers 37 numbers What algorithm? some algorithm

Slide 12

Slide 12 text

0 B B B B B B B B B B @ n1 . . . . . . . . . . . . n37 1 C C C C C C C C C C A 3 x 96 x 96 = 27648 numbers 37 numbers An early try: PCA + RF PCA Random Forest Regressor

Slide 13

Slide 13 text

No content

Slide 14

Slide 14 text

PCA + RF: Quick to run, good performance.

Slide 15

Slide 15 text

0 B B B B B B B B B B @ n1 . . . . . . . . . . . . n37 1 C C C C C C C C C C A 3 x 96 x 96 = 27648 numbers 37 numbers What algorithm? some algorithm

Slide 16

Slide 16 text

- “Analyzing complex image data? 
 Use a deep, convolutional neural network.” etc. lots of buzz over the past 2 years

Slide 17

Slide 17 text

What can they do? classification localization! ! …building block for 
 machine vision pipelines Russakovsky et al 2014 Karpathy et al 2014

Slide 18

Slide 18 text

What is a 
 deep convolutional neural net?

Slide 19

Slide 19 text

What is a 
 deep convolutional neural net? a (non-linear) function from to Rm Rn

Slide 20

Slide 20 text

What is a 
 deep convolutional neural net? a (non-linear) function from to Rm Rn

Slide 21

Slide 21 text

What is a 
 deep convolutional neural net? Self-learn a set of multi-channel convolutional kernels.
 (tiles, filter bank) Take advantage of approximate translational invariance. Fewer parameters to learn.

Slide 22

Slide 22 text

What is a 
 deep convolutional neural net? “Lots” of layers hierarchical / compositional

Slide 23

Slide 23 text

from “Visualizing and Understanding Convolutional Networks” Zeiler & Fergus 2013 A Hierarchy of Features

Slide 24

Slide 24 text

Why now?

Slide 25

Slide 25 text

• number of layers • width of layers • size of conv. kernels • pooling stride • choice of non-linear fn’s. • dropout • learning rates • … Lots of hyper-parameters ! ! ! ! ! ! ! ! Convnets might be overkill for a domain as “limited” as astronomical images. Computationally expensive to train (GPU-days or CPU-weeks) Cons

Slide 26

Slide 26 text

~All the best classifiers, including the best, 
 used convolutional neural nets. results …300+ teams

Slide 27

Slide 27 text

neuron 0 neuron 1 neuron 2 neuron 3 neuron 4 neuron 5 neuron 6 neuron 7 neuron 8 neuron 9 …

Slide 28

Slide 28 text

Astro. Applications? • objective (non-human) morphology • finding strong gravitational lenses • image-based redshifts (as opposed to photo-z.) • spectrograms or other 2d data • extending any of above to LSST-like scale • ?

Slide 29

Slide 29 text

“Smooth Galaxies” movie! 1000 nearby spirals 15 frames (galaxies) per second Subsequent frames chosen to be close in 
 morphology space and image space. http://vimeo.com/86254924