Demystifying Deep Learning for Data Scientists

on YouTube ↗01h 10m

" summary

Shows how to implement linear, logistic and feed‑forward networks from first principles using NumPy-style code and Jax automatic differentiation, with working notebooks (instructor and student) runnable on Binder. Covers explicit MSE and cross‑entropy formulas, manual gradient descent by hand and a training loop that fits a 41‑descriptor biodegradability model, with practical fixes (smaller step size, numeric stability). Claims a prior case of 100x RNN speedup with Jax and frames everything as the model‑loss‑optimizer workflow.

AI-written summary · about

" chapters (28)

1. 0:00Deep learning fundamentals
2. 1:26Workshop environment setup
3. 3:12Course goals prerequisites
4. 5:01Linear regression models
5. 8:27Loss functions for evaluation
6. 14:56Gradient descent optimization
7. 16:30Gradient-based optimization crash
8. 22:28Computational minimization with Python
9. 29:52Gradient descent by hand
10. 31:57Automatic differentiation with JAX
11. 33:28Gradient descent and linear regression
12. 34:52Model loss optimizer framework
13. 35:22Optimizer components breakdown
14. 37:17Linear regression in practice
15. 44:38Logistic regression models
16. 45:49Logistic curves exploration
17. 46:10Logistic regression parameters
18. 47:14Cross entropy loss
19. 49:02Logistic regression with optimizer
20. 54:37Neural networks approximators
21. 55:20Feed forward architecture
22. 56:43Biodegradable molecule dataset
23. 58:39Neural network from scratch
24. 1:00:13Training the model
25. 1:05:37Confusion matrix evaluation
26. 1:06:19Optimizer framework recap
27. 1:07:02Modifying model loss
28. 1:09:28Closing remarks resources

" related talks

• 44mMachine Learning ExposedJames Weaver · Codemotion Amsterdam 2017
• 51mMachine Learning Exposed Deep and Reinforcement LearningJames Weaver, Katharine Beaumont · Voxxed Days Vienna 2017
• 51mGetting Started with Deep Learning for Software DevelopersZoran Sevarac · Heapcon 2019
• 27mDemystifying Deep Neural NetworksRosie Campbell · Front-Trends 2017
• 35mDeconstructing Deep LearningMark West · Big Data Conference Europe 2019
• 51mDeep Learning - What is it and What It Can Do For YouDiogo Moitinho de Almeida · GOTO Copenhagen 2016
• 01h 04mWhat you always wanted to know about Deep Learning, but were afraid to askWei-Meng Lee · NDC London 2020
• 43m(Deep) Learn You a Neural Net for Great Good!Michael Stewart · PyData Los Angeles 2018