~/talk

Dive into Deep Learning with Scala

Talkat Scala Days Berlin 2018 · 2018-05-14..05-17· [machine-learning]
links:none on file
slides:none on file

Scala code uses TensorFlow for Scala to train on 800 labeled Scala Days photos and iterates from a simple fully connected net (~75–80% test accuracy) to a convolutional model (one 2D conv layer with 32 3×3 filters, bias, activation and max‑pool) running on GPU and reaching 95% test accuracy. The workflow shows concrete steps: dataset pipeline, cross‑entropy loss, gradient descent optimizer, batch training, and deployable model (live demo was flaky but screenshots confirm results).

AI-written summary · about

  1. 0:00Intro
  2. 0:33Electricity
  3. 1:08Deep Learning
  4. 1:40Getting Started
  5. 2:27Theres no free lunch
  6. 2:48Deep Learning in 3 Steps
  7. 3:09What is Deep Learning
  8. 3:54Computer Vision
  9. 4:45Speech Recognition
  10. 5:06Machine Translation
  11. 5:29What do these examples have in common
  12. 6:02Overlaying the examples
  13. 6:21Generic functions
  14. 6:48Models
  15. 7:02Classifiers
  16. 7:26Tensors
  17. 7:55Tensor Models
  18. 8:33Artificial Neuron
  19. 10:47Architecture
  20. 11:51Machine Learning
  21. 13:14Training Data Examples
  22. 14:00How Does Training Work
  23. 14:19First Step
  24. 15:04BackPropagation
  25. 15:33Gradient Descent
  26. 16:36Why Deep Learning Works
  27. 17:14Composition
  28. 18:28Narrow Intelligence
  29. 19:23Conclusion
  30. 19:38Why Scala
  31. 20:39Type Safety
  32. 21:37What features should a deep learning library provide
  33. 22:43Three deep learning libraries
  34. 24:25Example code
  35. 25:44Look for Existing Solutions
  36. 26:27Supervised Learning
  37. 27:46Data Source
  38. 28:51Architectures
  39. 30:01Convolutional
  40. 31:03Neural Network
  41. 31:37Model Building
  42. 34:19Training
  43. 35:19Training with more neurons
  44. 35:59convolutional layer
  45. 36:36network classifier
  46. 37:06run model
  47. 37:38deployment
  48. 38:31main takeaways
~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.