TO-DO.txt

Created; 15-03-2014 23:43
Edited;  23-03-2014 16:04

TO-DO List;

- compare old algorithm versus new algorithm with changes. Not versus other algorithms
- changes we will make: 
	- randomness less random for adding new antibodies
	- Different affinity-measures (like manhattan, sine/cosine, correlation (?)) -> sine/cosine needs to be implemented still
	- Different mutation process. Now; proportional to fitness. Comparison; what if based on static weight?
		o Global mutation process. Now; none. Improvement; mutate all cells such that there is a high probability that their affinity gets stronger/higher
		o [complex] Allow for self-adapting parameters for the thresholds
		o [complex] Implement the capability of the algorithm to maintain several distinct networks based on grouped affinities between cells. Requires first the identification of sub-networks inside the initial netwerk, followed by a split of the population into two new ones based on these sub-networks.

	* Determine evaluation function to use, or multiple
	* Test the chosen evaluation function(s)
	* Implement our improvements/comparisons
	* Create and additional evaluation class to evaluate the raw data into statisticall data