Scribd
Upload
19 views2 pages

Mod B Concise Summary

The document outlines key concepts in machine learning, including classification, linear and logistic regression, polynomial regression, gradient descent, regularization, dimensionality reduction, and PageRank. Each section provides definitions, equations, and relevant metrics or methods associated with the topic. The content serves as a concise reference for understanding fundamental machine learning algorithms and techniques.

Uploaded by

Jasvinder Kaur
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
19 views2 pages

Mod B Concise Summary

The document outlines key concepts in machine learning, including classification, linear and logistic regression, polynomial regression, gradient descent, regularization, dimensionality reduction, and PageRank. Each section provides definitions, equations, and relevant metrics or methods associated with the topic. The content serves as a concise reference for understanding fundamental machine learning algorithms and techniques.

Uploaded by

Jasvinder Kaur
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

1.

Classification

 Definition: Predict categorical class labels (e.g., spam or not spam).


 Algorithms: Decision Trees, k-NN, SVM, Naive Bayes.
 Metrics: Accuracy, Precision, Recall, F1-score, Confusion Matrix.

2. Linear Regression

 Definition: Predict continuous output (Y) from input features (X).


 Equation: Y=β0+β1X+ϵY = \beta_0 + \beta_1X + \epsilonY=β0+β1X+ϵ
 Assumptions: Linearity, Independence, Homoscedasticity, Normality.
 Metric: Mean Squared Error (MSE), R² score.

3. Logistic Regression

 Definition: Predict probability of binary outcomes (classification).


 Equation: P(Y=1)=11+e−(β0+β1X)P(Y=1) = \frac{1}{1 + e^{-(\beta_0 +
\beta_1X)}}P(Y=1)=1+e−(β0+β1X)1
 Output: Probability (0 to 1), apply threshold to classify.

4. Polynomial Regression

 Definition: Extends linear regression to model non-linear relationships.


 Equation: Y=β0+β1X+β2X2+...+βnXnY = \beta_0 + \beta_1X + \beta_2X^2 + ... +
\beta_nX^nY=β0+β1X+β2X2+...+βnXn
 Risk: Overfitting if degree is too high.

5. Gradient Descent

 Definition: Optimization algorithm to minimize loss function.


 Update Rule: θ=θ−α∂J∂θ\theta = \theta - \alpha \frac{\partial J}{\partial
\theta}θ=θ−α∂θ∂J
 Variants: Batch GD, Stochastic GD, Mini-batch GD.
 Learning Rate (α): Controls step size — too high = divergence; too low = slow.

6. Regularization

 Definition: Prevent overfitting by penalizing large coefficients.


 Types:
o L1 (Lasso): Shrinks some coefficients to zero (feature selection).
o L2 (Ridge): Shrinks coefficients but none to zero.
 Equation (Ridge): J=MSE+λ∑β2J = MSE + \lambda \sum \beta^2J=MSE+λ∑β2

7. Dimensionality Reduction

 Goal: Reduce features while preserving info.


 Methods:
o PCA (Principal Component Analysis): Projects data onto top variance
directions.
o t-SNE: For visualizing high-dim data in 2D/3D (not for prediction).
 Benefits: Faster models, removes multicollinearity, better visualization.

8. PageRank

 Definition: Algorithm by Google to rank web pages.


 Concept: A page is important if linked by other important pages.
 Equation (simplified):
PR(A)=(1−d)+d(∑PR(B)L(B))PR(A) = (1 - d) + d (\sum
\frac{PR(B)}{L(B)})PR(A)=(1−d)+d(∑L(B)PR(B))
where d is damping factor (~0.85), L(B) is number of outbound links from B.

You might also like