TD(0) Converges Provably Faster than the Residual Gradient Algorithm.
Ralf SchoknechtArtur MerkePublished in: ICML (2003)
Keyphrases
- experimental evaluation
- high accuracy
- dynamic programming
- learning algorithm
- optimization algorithm
- k means
- clustering method
- cost function
- similarity measure
- optimal solution
- computational cost
- computationally efficient
- detection algorithm
- matching algorithm
- times faster
- theoretical analysis
- linear programming
- expectation maximization
- search space
- computational complexity
- simulated annealing
- np hard
- significant improvement
- segmentation method
- tree structure
- improved algorithm
- highly efficient
- theoretical guarantees