A Gentle Introduction to Neural Networks and making your own with Python
summary
Clear, concrete outcomes: a 3-layer network implemented as a simple Python/Numpy pipeline and shared as a Python 3 notebook on GitHub and makeyourownneuralnetwork.blogspot.com. Code uses matrix multiplication for forward/backprop, random weight init, a tunable learning rate and epochs, trains on the MINST / EM mist handwriting set, reaches ~95% with plain code and jumps to ~98% after simple augmentation (rotations). Slides and source are linked for replication.
AI-written summary · about
chapters (32)
- 0:00Welcome
- 1:56Why neural networks matter
- 3:18Human vs computer strengths
- 5:32Linear models examples
- 8:04Training refines parameters
- 10:35Learning from data
- 14:10Learning rate moderation
- 14:31Boolean data classification
- 16:25Linear classifier limits
- 17:49From biological to artificial neurons
- 20:19Artificial neuron signal processing
- 22:04Neural network structure
- 23:07Where learning happens
- 24:56Resilience to damage
- 26:01Simplified calculations
- 27:03Node connections
- 28:56Matrix multiplication
- 31:05Error adjusts weights
- 33:53Gradient descent
- 38:16Building in Python
- 40:01Weight initialization
- 41:06Querying trained network
- 41:50Training function updates
- 42:12EMNIST handwritten digits
- 45:50Training randomness factors
- 47:13Data augmentation
- 47:56Reverse-engineering networks
- 49:21Live demo digits
- 50:31Scaling challenges
- 52:04Handling color images
- 53:50Scaling images
- 55:18Q&A
related talks
43mA Gentle Introduction to Neural Networks
43m(Deep) Learn You a Neural Net for Great Good!
06mNeural Networks Demystified: Part 2
58mNeural Networks: A Primer
01h 50mNeural Networks: Walkthrough
40mNeural Networks: Where do I start?!?
06mLearn Neural Networks With Go—Not Math! (Interview)
25mIntroduction to Neural Networks