ooogezil.com

About
A taxi on a street in Hong Kong.

A taxi on a street in Hong Kong.

Right of the divide: the same photograph through filter 20 of 64 in the first layer of ResNet-101, a network trained on ImageNet. It responds to red.

Courses

English中文

For beginners

  1. 1Machine learning: learning a function from examples
  2. 2Features: turning raw data into useful numbers
  3. 3Neural networks: functions that can bend
  4. 4Loss functions: measuring how wrong a model is
  5. 5Training models: walking down the loss map
  6. 6Evaluating models: will it work on new data?

Deep learning

  1. 7Deep networks: why depth helps
  2. 8Backpropagation: every gradient in one backward pass
  3. 9Regularization: fitting less to generalize better
  4. 10Convolutional networks: learning the filters
  5. 11Residual networks and normalization: training very deep networks
  6. 12Transformers: attention over a sequence

Agentic design

  1. 13Agents and workflows: from one model call to a system
  2. 14Tool use: letting the model act
  3. 15Reflection: checking and revising the model's work
  4. 16Planning: deciding the steps before taking them
  5. 17Memory and retrieval: what the agent knows, and when
  6. 18Multi-agent collaboration: teams of specialized agents