Methods for Feature Learning

• Start with a random dict and rep. Update D while keeping r fixed → find best r while keeping D fixed. Repeat until convergence.
• Move D in a direction to minimize loss and project it back
  • Gradient Descent or LinearRegression