Neural Networks: Walkthrough
summary
Step-by-step walkthrough of building and training simple neural nets using Python and TensorFlow with live demos: explains linear, binary threshold, sigmoid, tanh and ReLU is much faster than sigmoid activations, implements feedforward + backprop with gradient descent, shows a network learning a noisy sine/cosine function by ~50–100 iterations, highlights learning rate controls step size effects, covers over/underfitting and L1/L2 regularization, uses PCA for preprocessing and a small trained MNIST with CNN example (convolutional filters + pooling).
AI-written summary · about
chapters (12)
- 0:00Introduction to Neural Networks
- 4:22Biological and Artificial Neurons
- 9:20History of Neural Networks
- 15:48Linear Neuron Operations
- 22:42Activation Functions
- 33:51Network Structure
- 40:44Feed-Forward Process
- 49:46Training and Backpropagation
- 58:56Hyperparameters and Optimization
- 1:10:11Data Preparation
- 1:16:07Complex Architectures
- 1:18:45Conclusion and Resources
related talks
52mNeural Networks Walkthrough
43mA Gentle Introduction to Neural Networks
54mNeural networks by example
25mIntroduction to Neural Networks
51mMachine Learning Exposed Deep and Reinforcement Learning
40mNeural Networks: Where do I start?!?
06mNeural Networks Demystified: Part 2
43m(Deep) Learn You a Neural Net for Great Good!