Machine Learning (ML) Cheat Sheet

This document summarizes key concepts in Machine Learning for reference.

1. Learning Paradigms

Machine Learning can be mainly categorized based on learning methods as follows:

Supervised Learning

The model learns from Labeled Data, where the algorithm tries to find a Mapping Function from Input to Output.

• Classification: Predicting Categorical Labels (e.g., Spam Detection).
• Regression: Predicting Continuous Values (e.g., House Price Prediction).

Unsupervised Learning

The model learns from Unlabeled Data, where the algorithm tries to find Hidden Structure or patterns in the data.

• Clustering: Grouping Data Points with similar characteristics (e.g., K-Means).
• Dimensionality Reduction: Reducing the number of features while preserving as much information as possible (e.g., PCA).

Reinforcement Learning

An Agent learns to make decisions through actions in an Environment, receiving feedback in the form of Rewards or Penalties to find the Optimal Policy.

2. Bias-Variance Tradeoff

• Bias: Error resulting from using a model that is too simple, making it unable to capture the complexity of the data, leading to Underfitting.
• Variance: Error resulting from the model being too sensitive to fluctuations in the Training Data, causing it to capture noise in the data, leading to Overfitting.
• Tradeoff: Creating an effective model requires finding a sweet spot between Bias and Variance to achieve the Minimum Total Error on Unseen Data.

3. Evaluation Metrics

• Accuracy: The ratio of total correct predictions.
• Precision: The proportion of True Positives out of all Positive Predictions.
• Recall (Sensitivity): The proportion of True Positives out of all Actual Positives.
• F1 Score: The Harmonic Mean between Precision and Recall.