／var／logmarcus chiu

Machine Learning (ML) - Pattern Recognition

Created on · Last Modified on

Pattern Recognition

is the engineering application of various algorithms for the purpose of recognition of patterns in data

Machine Learning
  • constructs algorithms that can learn from data, identify patterns and make decisions
  • components of ML:
    • a pattern exists
    • the pattern cannot be pinned mathematically
    • data exists

ML - 4 Components

  • a computer program is said to learn from experience 𝐸 with respect to some class of tasks 𝑇 and performance measure 𝑃, if its performance at tasks in 𝑇, as measured by 𝑃, improves with experience 𝐸
  • a computer program is said to learn if its performance 𝑃 in task 𝑇 increases with experience 𝐸

Tasks (𝑇)

is the type of problem to solve

Model (𝑀)

Algorithm/Estimator/Predictor

is used to learn task 𝑇 (used over training-set)

Experience (𝐸)

is the type of training-set used by model 𝑀 (e.g. unsupervised learning, supervised learning, etc)

Performance Measure (𝑃)

is a method used to evaluate the trained model (used over test-set)

ML - 4 Components Example

Let’s decompose Linear Regression(i.e. linear regression is to find weights 𝜽 that reduce MSE over test-set when the algorithm experiences the training-set)

Task (𝑇)

to predict 𝑦 from 𝒙 by outputting 𝑦̂ = 𝜽𝑇𝒙

Model (𝑀)

Ordinary Least Squares (OLS) Regression is the algorithm used

Experience (𝐸)

a supervised set of (𝒙, 𝑦) tuples as training-set (labeled with 𝑦)

Performance Measure (𝑃)

Mean Square Error (MSE) over the test set

Resources