LPBoost - Algorithm

Algorithm

• Input:
  • Training set,
  • Training labels,
  • Convergence threshold
• Output:
  • Classification function

1. Initialization
   1. Weights, uniform
   2. Edge
   3. Hypothesis count
2. Iterate
   1. if then
      1. break
   2. solution of the LPBoost dual
   3. Lagrangian multipliers of solution to LPBoost dual problem

Note that if the convergence threshold is set to the solution obtained is the global optimal solution of the above linear program. In practice, is set to a small positive value in order obtain a good solution quickly.