A Gentle Introduction to Neural Networks
summary
A clear, hands-on intro that builds a tiny neural net in 30–40 lines of Python, uses sigmoid activation, trains with gradient descent and a tunable learning rate, and runs MNIST digit recognition reaching 70–80% baseline -> 98% with simple data augmentation. Slides, code and blog include the exact implementation; the demo runs on a laptop and can even run on a Raspberry Pi zero, successfully recognising noisy or partially damaged digits.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:25Questions
- 4:05Machine Learning
- 5:55What is Machine Learning
- 9:00Learning from Data
- 10:17Shifting the Line
- 12:13Adjusting the Line
- 14:15A Thought Process
- 18:08Two Lines
- 19:40Neural Networks
- 22:12Neural Network Architecture
- 31:25Code
- 35:28Results
- 39:12Live Demo
related talks
25mIntroduction to Neural Networks
56mA Gentle Introduction to Neural Networks and making your own with Python
01h 50mNeural Networks: Walkthrough
43m(Deep) Learn You a Neural Net for Great Good!
44mMachine Learning Exposed
01h 02mNeural Networks: A Primer
34mML from a software engineer's perspective
40mNeural Networks: Where do I start?!?