• -E,--weight <float> - take one argument, the penalty $\gamma$ for the HERD algorithm.
• -W,--window-size <integer> - size of the window $W$
• -H,--hmm <file> - the file in which is HMM
• -S,--sequence <file> - file containing input sequences, one sequence per line
• -m,--low-memory - if set, herd will use checkpointig technique to reduce memory requirements.
• -M,--high-memory - omit checkpointing technique
• -TO,--table-file-output <file> - store the precomputed posterior values to the file
• -TI,--table-file-input <file> - instead of computing the posterior values load them from file
• -F,--fancy-output - nice output
• -NF,--no-fancy-output - output as a sequence of labels
• -V,--viterbi - use viterbi algorithm insted of herd
• -B,--beam-width - Beam width for Viterbi algorithm (currently not implemented)
• -P,--parameters-file <file> - file containing the list of parameters
• -I,--hints-file <file> - file containing hints for the HERD algorithm to reduce running time
• -F,--hints-fuzz <integer> - the parameters for fuzzier hints.
• -A,--sample <file-prefix> <from> <to> - generate sequences and annotation from the model, save them
• -R,--precision <logoudble/logfloat/double/float> - which precision to use
• -O,--posterior - the posterior reward for the gain function
• -I,--incorrect-model - do not simplify the jpHMM

(:title Usage:)

Our implementation currently uses its own input and output formats.

The parameters:

• -E,--weight
• -W,--window-size
• -H,--hmm
• -S,--sequence
• -h,--help
• -m,--low-memory
• -M,--high-memory
• -TO,--table-file-output
• -TI,--table-file-input
• -F,--fancy-output
• -NF,--no-fancy-output
• -V,--viterbi
• -B,--beam-width
• -P,--parameters-file
• -I,--hints-file
• -F,--hints-fuzz
• -A,--sample
• -R,--precision
• -O,--posterior
• -R,--posterior-file
• -I,--incorrect-model