U ]<`m\@s*dZddlZz ejZWn4ek rJddlZedejddZYnXddZej dZ e e Z Gd d d Z d d Zd dZddZd4ddZdZd5ddZddZddZddZd6ddZddZd7d d!Zd"d#Zd$d%Zd&d'Zd(d)Zd*d+Zd,d-Zd.d/Z d0d1Z!d2d3Z"dS)8aA state-emitting MarkovModel. Note terminology similar to Manning and Schutze is used. Functions: train_bw Train a markov model using the Baum-Welch algorithm. train_visible Train a visible markov model using MLE. find_states Find the a state sequence that explains some observations. load Load a MarkovModel. save Save a MarkovModel. Classes: MarkovModel Holds the description of a markov model NzVFor optimal speed, please update to Numpy version 1.3 or later (current version is %s)cCsP||dkr|S||dkr |St||}|tt||t||S)z>Implement logaddexp method if Numpy version is older than 1.3.d)minnumpylogexp)ZlogxZlogyZminxyr.lib/python3.8/site-packages/Bio/MarkovModel.py logaddexp*s    r cCs@i}t|ddd}t|d}|D]\}}||||<q&|S)zAReturn a dictionary of values with their sequence offset as keys.N) enumeratelen)valuesdentriesnindexkeyrrr itemindex4s   rgYnc@s"eZdZdZdddZddZdS) MarkovModelz+Create a state-emitting MarkovModel object.NcCs"||_||_||_||_||_dS)zInitialize the class.N)statesalphabet p_initial p_transition p_emission)selfrrrrrrrr__init__Gs zMarkovModel.__init__cCs.ddlm}|}t|||d|S)z9Create a string representation of the MarkovModel object.r)StringIO)iorsaveseekread)rrhandlerrr__str__Qs    zMarkovModel.__str__)NNN)__name__ __module__ __qualname____doc__rr#rrrrrDs  rcCs&|}||s"td||f|S)zNRead the first line and evaluate that begisn with the correct start (PRIVATE).zI expected %r but got %r)readline startswith ValueError)r"startlinerrr_readline_and_check_start[s r-cCsrt|d}|dd}t|d}|dd}t||}t|t|}}t||_t|d}tt|D],}t|d||}t|d|j|<qrt||f|_ t|d}tt|D]>}t|d||}d d |ddD|j |ddf<qt||f|_ t|d }tt|D]@}t|d||}d d |ddD|j |ddf<q,|S) z.Parse a file handle into a MarkovModel object.zSTATES:r Nz ALPHABET:zINITIAL:z %s:r z TRANSITION:cSsg|] }t|qSrfloat.0vrrr |szload..z EMISSION:cSsg|] }t|qSrr.r0rrrr3s) r-splitrr rzerosrranger/rr)r"r,rrmmNMirrrloadcs,      * ,r;c Cs|j}|dd|j|dd|j|dtt|jD] }|d|j||j|fqD|dtt|jD]0}|d|j|ddd |j|Dfq||d tt|jD]0}|d|j|dd d |j|Dfqd S) z$Save MarkovModel object into handle.z STATES: %s  z ALPHABET: %s z INITIAL: z %s: %g z TRANSITION: z %s: %s css|]}t|VqdSNstrr1xrrr szsave..z EMISSION: css|]}t|VqdSr=r>r@rrrrBsN) writejoinrrr6r rrr)r7r"wr:rrrrs.rc st|t|}}|std|dk rDt|}|j|fkrDtd|dk rlt|}|j||fkrltd|dk rt|}|j||fkrtdg} t||D]} | fdd| Dqdd| D} t| d krtd t||| ||||d } | \} }}t ||| ||S) a Train a MarkovModel using the Baum-Welch algorithm. Train a MarkovModel using the Baum-Welch algorithm. states is a list of strings that describe the names of each state. alphabet is a list of objects that indicate the allowed outputs. training_data is a list of observations. Each observation is a list of objects from the alphabet. pseudo_initial, pseudo_transition, and pseudo_emission are optional parameters that you can use to assign pseudo-counts to different matrices. They should be matrices of the appropriate size that contain numbers to add to each parameter matrix, before normalization. update_fn is an optional callback that takes parameters (iteration, log_likelihood). It is called once per iteration. zNo training data given.N$pseudo_initial not shape len(states)5pseudo_transition not shape len(states) X len(states)5pseudo_emission not shape len(states) X len(alphabet)csg|] }|qSrrr@indexesrrr3sztrain_bw..cSsg|] }t|qSr)r r@rrrr3srz,I got training data with outputs of length 0)pseudo_initialpseudo_transitionpseudo_emission update_fn) r r*rasarrayshaperappendr _baum_welchr)rr training_datarKrLrMrNr8r9training_outputsoutputsZlengthsrArrrrrIrtrain_bwsB      rVic  Csd|dkrt|}n t||f}|dkr4t||f}nt|||f}|dkrXt||f}nt|||f}t|} t|} t|} |dk rt|} nd} |dk rt|}nd}|dk rt|}nd}d}ttD]j}t}|D] }|t|||| | | | || 7}q| dk r| |||dk r>t||dkr>qP|}qt dtdd| | | fDS)zfImplement the Baum-Welch algorithm to evaluate unknown parameters in the MarkovModel object (PRIVATE).Ng?z%HMM did not converge in %d iterationscSsg|]}t|qSr)rr)r1_rrrr3)sz_baum_welch..) _random_norm_copy_and_checkrrr6MAX_ITERATIONSLOG0_baum_welch_onefabs RuntimeError)r8r9rTrrrrKrLrMrN lp_initial lp_transition lp_emissionlpseudo_initiallpseudo_transitionlpseudo_emissionZ prev_llikr:ZllikrUrrrrRsV             rRc  Cst|} t|| ||||} t|| |||} t||| f} t| D]} || }t||f}t|D]R}t|D]D}| || ||||||| || d}||||<qpqd|t|| dddd| f<qBt|| f}t| D]0} t|D]"}t| |dd| f||| <qqt|}t|D]}t||ddf||<q2|dddf}|dk rt||}|t|}t|D]p}t|D],}t| ||ddf|||||<q|dk rt|||||<||t||||<qt|D]}t|t}t| D]:} || }t|D]"}t ||| ||| f||<q2q|t|}|dk rt|||}|t|}|||ddf<qt| dd| fS)zExecute one step for Baum-Welch algorithm (PRIVATE). Do one iteration of Baum-Welch based on a sequence of output. Changes the value for lp_initial, lp_transition and lp_emission in place. r Nr) r _forward _backwardrr5r6_logsum _logvecaddr[r )r8r9rUr_r`rarbrcrdTZfmatZbmatZlp_arctkZ lp_traverser:jlpZ lp_arcout_tZ lp_arcoutZksumrrrr\,s`         "       *    $    r\c Cst||df}||dddf<td|dD]n}||d}t|D]T} t} t|D]6} || |d|| | || |} t| | } qX| || |<qHq0|S)zImplement forward algorithm (PRIVATE). Calculate a Nx(T+1) matrix, where the last column is the total probability of the output. r Nrrr5r6r[r ) r8rir_r`rarUmatrixrjrkrllprobr:rmrrrres   ( rec Cst||df}t|dddD]j}||}t|D]T}t} t|D]6} || |d||| |||} t| | } qF| |||<q6q"|S)z'Implement backward algorithm (PRIVATE).r r rn) r8rir`rarUrorjrkr:rprlrmrrrrfs  ( rfcs&t|t|}}|dk r8t|}|j|fkr8td|dk r`t|}|j||fkr`td|dk rt|}|j||fkrtdgg}} t|t||D]P\} } t| t| krtd|fdd| D| fdd| Dqt||| ||||} | \} }}t||| ||S) a Train a visible MarkovModel using maximum likelihoood estimates for each of the parameters. Train a visible MarkovModel using maximum likelihoood estimates for each of the parameters. states is a list of strings that describe the names of each state. alphabet is a list of objects that indicate the allowed outputs. training_data is a list of (outputs, observed states) where outputs is a list of the emission from the alphabet, and observed states is a list of states from states. pseudo_initial, pseudo_transition, and pseudo_emission are optional parameters that you can use to assign pseudo-counts to different matrices. They should be matrices of the appropriate size that contain numbers to add to each parameter matrix. NrFrGrHzstates and outputs not alignedcsg|] }|qSrrr@)states_indexesrrr3sz!train_visible..csg|] }|qSrrr@)outputs_indexesrrr3s) r rrOrPr*rrQ_mler)rrrSrKrLrMr8r9training_statesrTZtoutputsZtstatesrArrrr)rrrqr train_visibles@       rucCst|}|r||}|D]}||dd7<qt|}t||f} |rV| |} |D]D}tt|dD].} || || d} } | | | fd7<qnqZtt| D]4} | | ddft| | ddf| | ddf<qt||f} |r| |} t||f} t||D]4\}}t||D]\}}| ||fd7<q&qtt| D]6} | | ddft| | ddf| | ddf<qV|| | fS)z.csg|]}j|qSr)rr@)r7rrr39s) r rrrrVERY_SMALL_NUMBERrrrr_viterbir6r) Z markov_modeloutputr8r_r`raresultsr:rscorer)rJr7r find_states#s   "rcCst|}g}t|D]}|dg|qt||f}||dd|df|dddf<td|D]r} || } t|D]\} |dd| df|dd| f|| | f} t| } | | d|| | f<| || | <qzqfg}g}t|dd|df} | D]&}||d|g|||dfq|r|\} }}| dkrT|||fn6||d| } | D] }|| d|g||fqhq&|S)zSImplement Viterbi algorithm to find most likely states for a given input (PRIVATE).Nrr )r r6rQrr5r|pop)r8r_r`rarriZ backtracer:ZscoresrjrkrlZi_scoresrJZ in_processrrrrrrr~=s4 $ 0$ "r~cCs~t|jdkr |tt|}nZt|jdkrrtt|D]4}||ddft||ddf||ddf<q:ntd|S)z"Normalize matrix object (PRIVATE).r Nz&I cannot handle matrixes of that shape)r rPr/rwr6r*)ror:rrrrvfs4rvcCst|}t|S)z%Normalize a uniform matrix (PRIVATE).)rrxrvrProrrr _uniform_normss rcCstj|}t|S)z$Normalize a random matrix (PRIVATE).)rrandomrvrrrrrXys rXcCstj|dd}|j|kr tdt|jdkrNtt|ddkrtdnPt|jdkrtt|D]*}tt||ddkrhtd|qhntd |S) zFCopy a matrix and check its dimension. Normalize at the end (PRIVATE).r )copyzIncorrect dimensiong?g{Gz?zmatrix not normalized to 1.0rzmatrix %d not normalized to 1.0z&I don't handle matrices > 2 dimensions)rZarrayrPr*r r]rwr6)roZ desired_shaper:rrrrYs  rYcCsFt|jdkr&t|t|jf}n|}t}|D]}t||}q2|S)z/Implement logsum for a matrix object (PRIVATE).r )r rPrZreshapeproductr[r )roZvecrwZnumrrrrgs rgcCsRt|t|kstdtt|}tt|D]}t||||||<q2|S)z5Implement a log sum for two vector objects (PRIVATE).zvectors aren't the same length)r AssertionErrorrr5r6r )Zlogvec1Zlogvec2Zsumvecr:rrrrhs rhcCst|}t|S)z-Return the exponential of a logsum (PRIVATE).)rgrr)Znumbersrwrrr _exp_logsumsr)NNNN)NNNNNNN)NNN)N)#r'rr AttributeErrorwarningswarn __version__rrZseedr}rr[rr-r;rrVrZrRr\rerfrursr|rr~rvrrXrYrgrhrrrrrsf    % E J^ ?. )