added (1+1)-EA experiments and documentation

0.8.69

added (1+1)-EA experiments

Major Change to the Experimentation API: Algorithms Instantiated per Run

We now instantiate the algorithms on per-run basis, no longer on a
per-problem basis.
The reason is that if we use the same instance for multiple runs, there
is a possibility that due to a programming issue, the runs are not
independent.
Such issues might be quite intricate.
Instantiation per-run is wasteful, indeed.
But the experimentation API is for experiments, not for productive use.
Thus, I will favor correctness and peace of mind over efficiency.

API Overhaul: Search Operators moved out of IBlackBoxProcess

The search operators have been moved out from IBlackBoxProcess.
While one can consider them as parts of the problem representation, one can as well consider them as part of the algorithm.
This has the advantage that the toString() method of the algorithm can return the full algorithm name without needing to know the black box configuration.

minor improvement of documentation plus one new test case

0.8.66

change in comments, added test case

more paco setups

0.8.65

more paco setups

PACO model now also stores edge set type in log files

PACO model now also stores edge set type in log files.
Since there are now two edge set types and the more suitable one is chosen automatically, this will provide better information about the experiment.

Further Speed-Up for PACO

Edges can now be added or removed from the matrix in O(1), if the number of nodes is low enough to keep a complete matrix in memory.

Hopefully Faster and Better PACO Implementation for the JSSP

The PACO implementation for the JSSP now in each step considers fewer possible nodes, namely only those associated with operations that can directly be executed. This might bring a speed-up for the model sampling by a factor of up to the number of machines.

minor bugfixes

• end time of runs now computed before synchronized block
• PACO heuristic values sum strictly monotonously increasing
• more PACO setups

First Experiment Draft Applying PACO to the JSSP

This is the first draft of the experiment with the population-based Ant Colony Optimization (PACO) implementation, which I try to apply to the Job Shop Scheduling Problem (JSSP).
The implementation and experiment probably are still severely flawed, but it is a first idea.