&h+dZddlZddlmZ ddlZddlm Z ejde Gdd Z d Z d Z d Zd ZdZdZdZdZdZ ddZedk(rddlmZedyy#e$rddlmZeddwxYw)z8Maximum Entropy code. Uses Improved Iterative Scaling. N)reduce)MissingPythonDependencyErrorzQPlease install NumPy if you want to use Bio.MaxEntropy. See http://www.numpy.org/)BiopythonDeprecationWarningzThe 'Bio.MaxEntropy' module is deprecated and will be removed in a future release of Biopython. Consider using scikit-learn instead.ceZdZdZdZy) MaxEntropyaHold information for a Maximum Entropy classifier. Members: classes List of the possible classes of data. alphas List of the weights for each feature. feature_fns List of the feature functions. Car data from example Naive Bayes Classifier example by Eric Meisner November 22, 2003 http://www.inf.u-szeged.hu/~ormandi/teaching >>> from Bio.MaxEntropy import train, classify >>> xcar = [ ... ['Red', 'Sports', 'Domestic'], ... ['Red', 'Sports', 'Domestic'], ... ['Red', 'Sports', 'Domestic'], ... ['Yellow', 'Sports', 'Domestic'], ... ['Yellow', 'Sports', 'Imported'], ... ['Yellow', 'SUV', 'Imported'], ... ['Yellow', 'SUV', 'Imported'], ... ['Yellow', 'SUV', 'Domestic'], ... ['Red', 'SUV', 'Imported'], ... ['Red', 'Sports', 'Imported']] >>> ycar = ['Yes','No','Yes','No','Yes','No','Yes','No','No','Yes'] Requires some rules or features >>> def udf1(ts, cl): ... return ts[0] != 'Red' ... >>> def udf2(ts, cl): ... return ts[1] != 'Sports' ... >>> def udf3(ts, cl): ... return ts[2] != 'Domestic' ... >>> user_functions = [udf1, udf2, udf3] # must be an iterable type >>> xe = train(xcar, ycar, user_functions) >>> for xv, yv in zip(xcar, ycar): ... xc = classify(xe, xv) ... print('Pred: %s gives %s y is %s' % (xv, xc, yv)) ... Pred: ['Red', 'Sports', 'Domestic'] gives No y is Yes Pred: ['Red', 'Sports', 'Domestic'] gives No y is No Pred: ['Red', 'Sports', 'Domestic'] gives No y is Yes Pred: ['Yellow', 'Sports', 'Domestic'] gives No y is No Pred: ['Yellow', 'Sports', 'Imported'] gives No y is Yes Pred: ['Yellow', 'SUV', 'Imported'] gives No y is No Pred: ['Yellow', 'SUV', 'Imported'] gives No y is Yes Pred: ['Yellow', 'SUV', 'Domestic'] gives No y is No Pred: ['Red', 'SUV', 'Imported'] gives No y is No Pred: ['Red', 'Sports', 'Imported'] gives No y is Yes c.g|_g|_g|_y)zInitialize the class.N)classesalphas feature_fns)selfs ]/mounts/lovelace/software/anaconda3/envs/py312/lib/python3.12/site-packages/Bio/MaxEntropy.py__init__zMaxEntropy.__init__Zs  N)__name__ __module__ __qualname____doc__rrr rr$s 3jrrcg}t|jt|jk(sJ|jD]L}d}t |j|jD]\}}|||||zz }|j |N|S)zCalculate the log of the probability for each class. me is a MaxEntropy object that has been trained. observation is a vector representing the observed data. The return value is a list of unnormalized log probabilities for each class. )lenr r r zipappend)me observationscoresklasslprobfnalphas r calculater!asF r~~ #bii. 00R^^RYY7 4IB R U+e3 3E 4 e  Mrct||}|d|jd}}tdt|D]}|||kDs |||j|}}!|S)z%Classify an observation into a class.r)r!r ranger)rrr max_scoreris r classifyr'rse r; 'Fay"**Q-uI 1c&k "8 !9y %ay"**Q-uI8 Lrci}tt|D]7}tt|D]}|||||}|dk7s||||f< 9|S)a>Evaluate a feature function on every instance of the training set and class (PRIVATE). fn is a callback function that takes two parameters: a training instance and a class. Return a dictionary of (training set index, class index) -> non-zero value. Values of 0 are not stored in the dictionary. r)r$r)rxsr valuesr&jfs r _eval_feature_fnr-|sgF 3r7^#s7|$ #A2a5'!*%AAv!"1v ## Mrci}t|D] \}}|||< |Dcgc]}|| }}g} t|} |D]B} d} t| D]} | | j| || fdz } | j | | z D| Scc}w)zCalculate the expectation of each function from the data (PRIVATE). This is the constraint for the maximum entropy distribution. Return a list of expectations, parallel to the list of features. r) enumeraterr$getr)r)ysr features class2indexindexkeyyys_iexpectNfeaturesr&s r _calc_empirical_expectsr<sK(! s C!$& 'qKN 'D ' F BA q .A aa\1- -A . a!e  M (s Bct||||}g}|D]J}d}|jD]\\}} } |||| | zz }|j|t|z L|S)zCalculate the expectation of each feature from the model (PRIVATE). This is not used in maximum entropy training, but provides a good function for debugging. r)_calc_p_class_given_xitemsrr) r)r r2r p_yxexpectsr:sumr&r+r,s r _calc_model_expectsrCs{ !Wh ?DG&  "IFQA 471:> !C "sSW}% & Nrctjt|t|f}t|t|k(sJt||D]3\}}|j D]\\}}} |||xx|| zz cc<5tj |}t t|D]} t|| } || | z || <|S)zCalculate conditional probability P(y|x) (PRIVATE). y is the class and x is an instance from the training set. Return a XSxCLASSES matrix of probabilities. )npzerosrrr?expr$rB) r)r r2r prob_yxr:r xr6r,r&zs r r>r>s hhBW./G x=CK ''h/'  'IFQA AJqMUQY &M ''ffWoG 3r7^$  OQZ!^ $ Nrctj||f}|D]-}|jD]\\}}}|||xx|z cc</|S)z/Calculate a matrix of f sharp values (PRIVATE).)rErFr?)r9nclassesr2f_sharpr:r&r+r,s r _calc_f_sharprNs[hh8}%G  IFQA AJqMQ M  Nrc^d}d}||krdx} } |jD]E\\} } } || | | ztj||| | zz}| |z } | ||| | zz } G|| |z z | |z } } | | z }||z}tj||kr |S|dz}||krt d)z,Solve delta using Newton's method (PRIVATE).rrr#z Newton's method did not converge)r?rErGfabs RuntimeError)r9r:rM empiricalrHmax_newton_iterationsnewton_convergedeltaitersf_newton df_newtonr&r+r,prodratios r _iis_solve_deltar[s E E ' '""9  .IFQA1:a=1$rvvegajm.C'DDD  H  1 - -I .((Q,6 Q)9$  775>O +  L  ' '=>>rc t||||}t|} |dd} tt|D]&} t| || ||| |||} | | xx| z cc<(| S)zBDo one iteration of hill climbing to find better alphas (PRIVATE).N)r>rr$r[) r)r r2rMr e_empiricalrSrTr@r9 newalphasr&rUs r _train_iisr_s} !Wh ?D BAq I 3v;   QK  N  !   !    rc |s tdt|t|k7r td||} }tt|} |D cgc]} t | || } } t t|t| | } t || | | }dgt| z}d}||krt|| | | ||||}t||Dcgc]\}}tj||z }}}ttj|d}|}t}|| |c|_|_|_|||||kr |S||krt#dcc} wcc}}w)aFTrain a maximum entropy classifier, returns MaxEntropy object. Train a maximum entropy classifier on a training set. training_set is a list of observations. results is a list of the class assignments for each observation. feature_fns is a list of the features. These are callback functions that take an observation and class and return a 1 or 0. update_fn is a callback function that is called at each training iteration. It is passed a MaxEntropy object that encapsulates the current state of the training. The maximum number of iterations and the convergence criterion for IIS are given by max_iis_iterations and iis_converge, respectively, while max_newton_iterations and newton_converge are the maximum number of iterations and the convergence criterion for Newton's method. zNo data in the training set.z2training_set and results should be parallel lists.rrzIIS did not converge) ValueErrorrsortedsetr-rNr<r_rrErPraddrr r r rQ) training_setresultsr update_fnmax_iis_iterations iis_convergerSrTr)r1r rr2rMr]r rVnalphasrIr6diffrs r trainrls|4 788 <CL(MNN7BS\"GGRR \7;RHRC -s7|XFG*"b'8DKUS] "F E $ $       !   ,/vw+?@41aA@@bffdA& \06- 2:r~  bM ,   I3 $ $.122CS*As E "E__main__) run_doctest)verbose)Ni'gh㈵>dg|=)rwarnings functoolsrnumpyrE ImportErrorBiorrwarnrr!r'r-r<rCr>rNr[r_rlr Bio._utilsrnrrr rxs ,  A::z"".&,2HIX z&U 0 & $  s A//B