## Branch & Bound, Graphical Models and Structured Grids patterns

Escrito por gacevedo el 12 de noviembre de 2009#### Branch & Bound

This pattern is used for problems where you have a very large space and you need to search to make a decision or find an optimal solution. Of course the space is too large for enumerating every point.

In order to find the solution to the problem, this pattern has four operations: Branching, Evaluation, Bounding and Pruning.

This kind of problems are good for parallelism, even thou synchronization is complicated. You will want to reduce communication and increase computation.

#### Graphical Models

This pattern gave me a bad impression, because I thought I was going to read about graphics rather than statistics and probabilities. Also as some classmates mentioned, it lacked of a good and clear example.

#### Structured Grids

This pattern was better written I think, at least it had several examples (thou I missed source code!). The main idea is to decompose (using geometric decomposition) the problem into smaller arrays and perform operations over the arrays in parallel.

The double buffering and adaptive mesh refinement idea was very interesting to me and seems to be very useful for image processing.