Given a loaded data set
generate a translate template with 1 record for each field
fields in the *.dfn file named "ignore" will have template usage = IGNORE
fields marked as "output" it will have template usage = OUTPUT and an output
template will be generated
continuous fields are scaled to between 0.1 and 0.9 (using min/mean/max)
provide an input translate template for this field
if the field is a regular input field, scale to 0.0 -- 1.0
if it is a "predicted" field , scale to 0.1 to 0.9
provide an input translate template for this field
if the field is an input field, create a one of N code
if it is an output field, create a one of N code
if it is marked ignore, then ignore it
if field only contains 2 discrete values, treat it as a single binary template
get the internal config for either
NNTRAIN = training from an external data file
NNTEST = testing (with network locked) from an external data file
or NNRUN = running (with network locked) from data placed in the input buffer
get the internal config for either
NNTRAIN = training from an external data file
NNTEST = testing (with network locked) from an external data file
or NNRUN = running (with network locked) from data placed in the input buffer
get the internal config for either
NNTRAIN = training from an external data file
NNTEST = testing (with network locked) from an external data file
or NNRUN = running (with network locked) from data placed in the input buffer
get the internal config for either
NNTRAIN = training from an external data file
NNTEST = testing (with network locked) from an external data file
or NNRUN = running (with network locked) from data placed in the input buffer
read the next n records from a text file into a vector of String arrays
assumes that the variables/ field list have already been created
that the textData and numericData vectors (of arrays) have been allocated
and that the BufferedReader is initialized
wraps at end of file
return a hashtable of operator names (keys) and fitness values (Doubles)
fitness values must sum to 100
Note: this hashtable may be extended to using operator fitness objects vs fixed values
so that we can adapt the operators selection during the search process
return a hashtable of operator names (keys) and fitness values (Doubles)
fitness values must sum to 100
Note: this hashtable may be extended to using operator fitness objects vs fixed values
so that we can adapt the operators selection during the search process
return the vocabulary used for the chromosomes
Note: this is usually the alphabet "01" or "abc", but could be
a more descriptive string "real" or "hybrid" if a non-String chromosome
format is used.