Gradient Descent

Source folder: Gradient-Descent/

Summary

Homework project focused on understanding gradient descent behavior through custom objective functions and visual experimentation.

Analysis Used

  • Construction of scalar objective functions and analytical derivatives.
  • Step-by-step 1D gradient descent implementation and testing.
  • Experiments with learning-rate and initialization choices.
  • Extension to higher-dimensional gradients and contour analysis.
  • Visual interpretation of convergence versus oscillation/divergence.

Technologies and Methods

  • Python, Jupyter Notebook
  • numpy
  • plotly.graph_objects
  • IPython.display