Dynamic Programming
Table of Contents:
Applications
Dynamic Programming (DP) …
Linux utility diff
,
Fibonacci Numbers
Least Common Subsequence
Longest Increasing Subsequence
Longest Common Subsequence
Knapsack Problem
Break into house, and take items of most value
Solving Recurrences
Integral Images
SURF, Viola Jones rely upon fast computation of Integral Images for fast local feature detection and object detection. Dynamic Programming
Max-Pooling in ConvNets
[1]
Graphical Models
Koller and Friedman [2] use DP extensively. Variable Elimination (pp. 292-296, 337) for the summation of the joint distribution from the inside out, rather from the outside in, and cache the intermediate results to avoid repeated computation.
DP also arises in Belief Propagation (p. 356), Clique Tree Queries (p. 371), MAP and marginal MAP queries (p. 596).
References
[1]. Alessandro Giusti, Dan C. Cires¸an, Jonathan Masci, Luca M. Gambardella, Jurgen Schmidhuber. Fast Image Scanning with Deep Max-Pooling Convolutional Neural Networks. PDF.
[2]. Daphne Koller and Nir Friedman. Probabilistic Graphical Models: Principles and Techniques. 2009.