this is the fourth release of the Iedera software : this command
lined tool allows to design and select subset seeds according to 
Bernoulli, Markov or any HMM model.

Version 1.00
============
- Fork of the Hedera program, but without any nucleotide seed
- Vectorized subset seed model proposed
- Hopcroft algorithm has been modified : the old version was in fact 
not really carefully programmed (and thus worked in more than n.log(n)).

Version 1.01
============
- new subset seed automaton implementation used (CIAA07 algorithm).
- possibility to index sequence on "one in cycle'" positions : phase for
multiple seeds is optimized. This is the "-c" parameter
- hill climbing heuristic (-k command)
- keep (minimized) automata version of seeds that remain unchanged (faster).
- [ homogeneous model added (not command lined yet) ]
- [ input/output of automata added (not command lined yet) ]
- bug solved in the PARSEMATRIX column count
- pareto set file input/output (to merge previously computed results)
- bug solved on complete enumeration (memory leak)

Version 1.02
============
- "homogeneous model" modified and command-lined
- add "-transitive" and "-spaced" command-line parameters to provide 
  "nucleic seeds" 
- add "-BSymbols

Version 1.03
============
- "-mx" exclude pattern parameter added
- "-MF" and "-MS" files parameters ... (command line limits)
- Lossless seeds added (work in progress)
- Cycle algorithm modified (now phase is always fixed ... it is a more
usefull case) with (for each seed) independent cycles and number of seeds
inside the cycle.
- Cycle algorithm command lined for the "-m" option
- Homogeneous algorithm and parameters modified (no max score, now the
score of the full length alignment must be >= than any subpart of it)

Version 1.04
============
- "-fF" option added to input the probabilistic automaton used to estimate the sensitivity. 
  This parameter can also be used in lossless mode :
  -> In lossless mode, if some final states are present in this automaton, they are considered as 
     "matching" ones independently of the seed chosen, thus "increasing" in some way
     the language recognized by the seed ... they can thus be considered as "reject" states since
     they are alway matching (note that the lossless algorithm only focus on non final states). 
  -> In lossy mode, final states are excluded from the language being recognized.
- Bug on the hillclimbing process solved (but some other may appear :op ).
- Seed positions are now also optimized in the hillclimbing heuristic.
- Symetric spaced seeds option added ([not a very efficient programming, but usefull anyway...])


Version 1.05
============
- Seed positions optimized without "self-coverage" between positions 
- multihit "-y" parameter added (it computes the multihit criterion, with overlap between seeds), 
  (Hopcroft minimization now accepts differently labelled final states).

Version 1.05b
-------------
- a bug on positioned seeds has been solved (in some cases, not all positions were considered)
- a improvement on positioned seeds (during hillclimbing) has been set.
