Hi👋

I’m Renyi Qu. Welcome to my minimalist AI handbook.

Each model/method is decomposed in the following structure:

• Why: Why do we need it? What’s the motivation? What’s the problem?
• What: What’s the intuition? What’s the mechanism? What are the components?
• How: How do we use it? How do we apply it?
• When: When can we use it? What are the assumptions/conditions?
• Where: Where can we apply it?
• Pros & Cons: What should we consider when using it?

Common notations include (section-specific notations take higher priority):

• : scalar / concept abstraction
• : vector
• : matrix / random variable
• : set
• : norm (for continuous vectors) / count (for discrete vectors)
• : count
• : estimator
• : #samples in the input batch
• : #features in the input sample
• : #classes in the training set
• : sample index
• : feature index
• : class index
• : training set
• : class set
• : all values of th feature for samples from th class
• : input matrix of shape (add if bias is needed)
• : output vector of shape
• : params (add if bias is needed)