06_loss_landscape

📉 Loss Landscape Visualization

This project visualizes how the loss function (MSE) changes with different weight and bias combinations in a simple linear regression task.

---

🎯 Problem Statement

In deep learning, understanding how model parameters affect the loss surface is crucial. This visualization helps interpret:

• Where the global minima lies
• How gradient descent navigates the surface
• The shape of convex vs. non-convex loss functions

---

🔧 Tools & Libraries Used

• Python 3
• NumPy
• Matplotlib
• mpl_toolkits.mplot3d for 3D plotting

---

📊 What It Does

• Generates a toy linear dataset: ( y = 2x + 3 + \epsilon )
• Computes the Mean Squared Error (MSE) over a grid of weight (w) and bias (b) values
• Plots:
  • A 2D Contour Plot of the loss surface
  • A 3D Surface Plot of the same

---

📈 Sample Output

Contour Plot	3D Surface

▶️ How to Run

pip install numpy matplotlib
python loss_surface.py