Definition: Learning from labeled data to predict outcomes for new, unseen data.
varsha-sweetie / ML
May 20, 2025 13:40
Missing something?
ML
A concise cheat sheet outlining the key concepts, algorithms, and differences between supervised and unsupervised learning methods in machine learning.
Core Concepts
Supervised Learning
|
|
|
Goal: To map input variables (X) to an output variable (Y) based on a labeled dataset. |
|
Process: Model trained on labeled data → Model predicts outcome on new data → Accuracy measured by comparing predicted vs. actual values. |
|
Common Use Cases: Classification (categorizing data) and Regression (predicting continuous values). |
Unsupervised Learning
|
Definition: Learning from unlabeled data to discover hidden patterns or structures. |
|
Goal: To find inherent structure in data without explicit labels. |
|
Process: Model explores unlabeled data → Identifies patterns, clusters, or reduces dimensionality. |
|
Common Use Cases: Clustering (grouping similar data points) and Dimensionality Reduction (reducing the number of variables). |
Supervised Learning Algorithms
Classification Algorithms
|
Logistic Regression: Predicts the probability of a binary outcome. Good for binary classification problems. |
|
Support Vector Machines (SVM): Finds the optimal hyperplane that separates data into classes. Effective in high dimensional spaces. |
|
Decision Trees: Tree-like structure to classify data based on features. Easy to interpret, but prone to overfitting. |
|
Random Forest: Ensemble of decision trees that improves accuracy and reduces overfitting. |
|
Naive Bayes: Applies Bayes’ theorem with strong (naive) independence assumptions between features. Simple and fast, but assumptions may not hold in real-world data. |
|
K-Nearest Neighbors (KNN): Classifies a data point based on the majority class of its k-nearest neighbors. |
|
Confusion Matrix: A table that summarizes the performance of a classification model by showing the counts of true positive, true negative, false positive, and false negative predictions. Used to calculate various metrics like accuracy, precision, recall, and F1-score. |
|
Stochastic Gradient Descent (SGD): An iterative optimization algorithm used to find the minimum of a cost function. It updates model parameters using the gradient of the cost function computed on a small, randomly selected subset of the training data (a mini-batch) at each iteration. SGD is computationally efficient and suitable for large datasets. |
Regression Algorithms
|
Linear Regression: Models the relationship between variables using a linear equation. Simple and interpretable. |
|
Polynomial Regression: Models the relationship using a polynomial equation. Can capture non-linear relationships. |
|
Support Vector Regression (SVR): Uses SVM principles to predict continuous values. |
|
Decision Tree Regression: Uses decision trees to predict continuous values. Prone to overfitting. |
|
Random Forest Regression: Ensemble of decision trees for regression. Improves accuracy and reduces overfitting. |