Linear Regression
Introduction
Linear regression is one of the simplest and most fundamental algorithms in machine learning. It models the relationship between variables by fitting a linear equation to observed data.
The Mathematics
For simple linear regression with one input variable, we model:
y = mx + b
Where:
- y: predicted output
- x: input feature
- m: slope (weight)
- b: y-intercept (bias)
We find the best parameters by minimizing the Mean Squared Error (MSE):
MSE = (1/n) × Σ(yᵢ - ŷᵢ)²
Interactive Linear Regression
Click to add data points, then train the model using gradient descent or fit analytically:
Click to add data points
Controls
Model Parameters
Slope (m) = 0.000
Intercept (b) = 0.000
MSE Loss = 0.0000
Tips:
- Red lines show residuals (errors)
- Gradient descent iteratively improves the fit
- Analytical solution gives the optimal fit instantly
Gradient Descent vs Analytical Solution
Gradient Descent
- Iterative optimization algorithm
- Works for any differentiable loss function
- Scales to large datasets and complex models
- Foundation for training neural networks
- Requires tuning learning rate
Analytical Solution
- Closed-form solution using calculus
- Gives optimal parameters directly
- Only works for linear regression
- Computationally expensive for large datasets
- No hyperparameters to tune
Key Takeaways
- Linear regression finds the best-fitting line through data points
- The goal is to minimize the squared errors between predictions and actual values
- Gradient descent is a general optimization technique that works for many ML algorithms
- Understanding linear regression helps grasp concepts like loss functions, optimization, and overfitting
- It's the foundation for more complex algorithms like neural networks