v0.9.2

Added

• (python) parallel_breeding for De (#296)
• (site) example families as tabs, and a sidebar of every example (#279)
• (benchmarks) single-objective scenarios only, for now (the benchmark harness's commands changed; the library's API didn't) (#283)
• (benchmarks) a matched suite of three problems, one method each (a new harness command set, as above) (#284)
• breed a GA's offspring in parallel with GaBuilder::parallel_breeding (#289)
• parameter control with Engine::control, and re-evaluation in the single-objective algorithms (#292)
• isolated islands that never migrate, with Topology::Isolated (#293)
• (python) parameter control and re-evaluation with run(control=...) (#294)
• build DE trials and ES offspring in parallel with parallel_breeding (#295)
• keep what a fitness function computed alongside the fitness (#301)

Performance

• push each child straight into the offspring when breeding serially (#290)
• (python) twice as fast on bit genomes, one GIL attach per generation (#291)
• take an ES's step-size logarithms once per generation (#298)

Documentation

• the roadmap as released: this release's control and parallel breeding, genetic programming in 0.10 (#299)