In the Name of Whiskey

Using Machine Learning and TensorFlow to make sense of scotches. Presented at Ruby on Ales 2016.

Transcript

1. 1 @juliaferraioli In the Name of Whiskey Julia Ferraioli Software Engineer @juliaferraioli
2. 2 @juliaferraioli @juliaferraioli Engineer @ Google I’ve done some stuff in another life…
3. 3 no PhD ⇒ no more machine learning for me
4. 4 wait... what build rule checks for a PhD?
5. 5 → ⚙ → ❓ data algorithm answer
6. 6 E_TOO_FEW_HUGS
7. 7 can I hug that?
8. 8 @juliaferraioli Can I hug that?
9. 9 @juliaferraioli Can I hug that?
10. 10 but what about the whiskey?
11. 11 • A deep, abiding love for whiskey • Some manageable, objective data • Enthusiasm for math… • ...or making computers do math What do we have?
12. 12 let’s talk about data
13. 13 @juliaferraioli (x,y) Feature vectors
14. 14 @juliaferraioli (x,y,z) Feature vectors
15. 15 @juliaferraioli (x,y,z,?,?,?,?,...) Feature vectors
16. 16 our data
17. 17 @juliaferraioli Distillery → Balvenie Body → 3 Notes → Honey:3, Floral:2, ... Region → Speyside Lat / Lon → Lat:57.4527074, Lon:-3.1237258 Data example
18. 18 @juliaferraioli [Name, Region, Body, Notes 0 ,...,Notes n , Lat, Lng] Our potential feature vector
19. 19 @juliaferraioli [5,3,2,1,0,0,3,2,1,0,2,2,2,57.4527074,-3.1237258] Our potential feature vector
20. 20 Operates over tensors: n-dimensional arrays Using a flow graph: data flow computation framework A quick look at TensorFlow • Intuitive construction • Fast execution • Train on CPUs, GPUs • Run wherever you like
21. 21 @juliaferraioli import tensorflow as tf sess = tf.InteractiveSession() # don’t mess with passing around a session whiskey_is_fun = tf.constant([6.2, 12.0, 5.9], shape = [1, 3]) beer_is_ok_too = tf.constant([9.3, 1.7, 8.8], shape = [3, 1]) matrices_omg = tf.matmul(whiskey_is_fun, beer_is_ok_too) print(matrices_omg.eval()) # => [[ 129.97999573]] sess.close() # let’s be responsible about this What does TensorFlow code look like?
23. 23 k-means clustering
24. 24 @juliaferraioli k = # of clusters
25. 25 @juliaferraioli pick k random n-dimensional cluster centers while not converged and num_steps < max_steps: assign points to nearest cluster center update cluster centers to be the mean of the points assigned to it return cluster centers, memberships The algorithm
26. to the container!
28. 28 feed forward neural networks [with hidden layer(s)]
29. 29 @juliaferraioli ["this", "movie", "was", "great"] [" "] Input → Output → ????? →
30. 30 @juliaferraioli ["this", "movie", "was", "great"] ["POS"] Input → Hidden → Output →
31. 31 @juliaferraioli initialize hidden, output layers (weights and biases) define the training step (optimization) while num_steps > limit: grab next set of training data and labels perform training step evaluate on test data The algorithm (simplified)
32. to the container!
34. 34 lessons
35. I’ve been pronouncing all of my scotch incorrectly
36. data collection is the hard part
37. common sense is not my strong suit
38. thinking in n-dimensions makes my brain hurt
39. NEED MOAR WHISKEY!
40. NEED MOAR DATA!
41. 41 @juliaferraioli Resources • TensorFlow: http://bit.ly/tensorflow-oss • k-means Clustering Single Malts: http://bit.ly/k-means-scotch • k-means in TensorFlow: http://bit.ly/k-means-tensorflow • Feed forward neural networks in TensorFlow: http://bit.ly/ff-nn • Learning TensorFlow: http://bit.ly/learn-tensorflow • Gentle Introduction to ML: http://bit.ly/gentle-intro-ml
42. Octopus: image by Feans thanks! Toolbox: image by Tanemori