First commit

libsvm-3.36/python/MANIFEST.in

@@ -0,0 +1,2 @@
+include  cpp-source/*
+include  cpp-source/*/*

libsvm-3.36/python/Makefile

@@ -0,0 +1,4 @@
+all = lib
+
+lib:
+	make -C .. lib

libsvm-3.36/python/README

@@ -0,0 +1,511 @@
+----------------------------------
+--- Python interface of LIBSVM ---
+----------------------------------
+
+Table of Contents
+=================
+
+- Introduction
+- Installation via PyPI
+- Installation via Sources
+- Quick Start
+- Quick Start with Scipy
+- Design Description
+- Data Structures
+- Utility Functions
+- Additional Information
+
+Introduction
+============
+
+Python (http://www.python.org/) is a programming language suitable for rapid
+development. This tool provides a simple Python interface to LIBSVM, a library
+for support vector machines (http://www.csie.ntu.edu.tw/~cjlin/libsvm). The
+interface is very easy to use as the usage is the same as that of LIBSVM. The
+interface is developed with the built-in Python library "ctypes."
+
+Installation via PyPI
+=====================
+
+To install the interface from PyPI, execute the following command:
+
+> pip install -U libsvm-official
+
+Installation via Sources
+========================
+
+Alternatively, you may install the interface from sources by
+generating the LIBSVM shared library.
+
+Depending on your use cases, you can choose between local-directory
+and system-wide installation.
+
+- Local-directory installation:
+
+    On Unix systems, type
+
+    > make
+
+    This generates a .so file in the LIBSVM main directory and you
+    can run the interface in the current python directory.
+
+    For Windows, the shared library libsvm.dll is ready in the
+    directory `..\windows' and you can directly run the interface in
+    the current python directory. You can copy libsvm.dll to the
+    system directory (e.g., `C:\WINDOWS\system32\') to make it
+    system-widely available. To regenerate libsvm.dll, please
+    follow the instruction of building Windows binaries in LIBSVM
+    README.
+
+- System-wide installation:
+
+    Type
+
+    > pip install -e .
+
+    or
+
+    > pip install --user -e .
+
+    The option --user would install the package in the home directory
+    instead of the system directory, and thus does not require the
+    root privilege.
+
+    Please note that you must keep the sources after the installation.
+
+    For Windows, to run the above command, Microsoft Visual C++ and
+    other tools are needed.
+
+    In addition, DON'T use the following FAILED commands
+
+    > python setup.py install (failed to run at the python directory)
+    > pip install .
+
+Quick Start
+===========
+
+"Quick Start with Scipy" is in the next section.
+
+There are two levels of usage. The high-level one uses utility
+functions in svmutil.py and commonutil.py (shared with LIBLINEAR and
+imported by svmutil.py). The usage is the same as the LIBSVM MATLAB
+interface.
+
+>>> from libsvm.svmutil import *
+# Read data in LIBSVM format
+>>> y, x = svm_read_problem('../heart_scale')
+>>> m = svm_train(y[:200], x[:200], '-c 4')
+>>> p_label, p_acc, p_val = svm_predict(y[200:], x[200:], m)
+
+# Construct problem in python format
+# Dense data
+>>> y, x = [1,-1], [[1,0,1], [-1,0,-1]]
+# Sparse data
+>>> y, x = [1,-1], [{1:1, 3:1}, {1:-1,3:-1}]
+>>> prob  = svm_problem(y, x)
+>>> param = svm_parameter('-t 0 -c 4 -b 1')
+>>> m = svm_train(prob, param)
+
+# Precomputed kernel data (-t 4)
+# Dense data
+>>> y, x = [1,-1], [[1, 2, -2], [2, -2, 2]]
+# Sparse data
+>>> y, x = [1,-1], [{0:1, 1:2, 2:-2}, {0:2, 1:-2, 2:2}]
+# isKernel=True must be set for precomputed kernel
+>>> prob  = svm_problem(y, x, isKernel=True)
+>>> param = svm_parameter('-t 4 -c 4 -b 1')
+>>> m = svm_train(prob, param)
+# For the format of precomputed kernel, please read LIBSVM README.
+
+
+# Other utility functions
+>>> svm_save_model('heart_scale.model', m)
+>>> m = svm_load_model('heart_scale.model')
+>>> p_label, p_acc, p_val = svm_predict(y, x, m, '-b 1')
+>>> ACC, MSE, SCC = evaluations(y, p_label)
+
+# Getting online help
+>>> help(svm_train)
+
+The low-level use directly calls C interfaces imported by svm.py. Note that
+all arguments and return values are in ctypes format. You need to handle them
+carefully.
+
+>>> from libsvm.svm import *
+>>> prob = svm_problem([1,-1], [{1:1, 3:1}, {1:-1,3:-1}])
+>>> param = svm_parameter('-c 4')
+>>> m = libsvm.svm_train(prob, param) # m is a ctype pointer to an svm_model
+# Convert a Python-format instance to svm_nodearray, a ctypes structure
+>>> x0, max_idx = gen_svm_nodearray({1:1, 3:1})
+>>> label = libsvm.svm_predict(m, x0)
+
+Quick Start with Scipy
+======================
+
+Make sure you have Scipy installed to proceed in this section.
+If numba (http://numba.pydata.org) is installed, some operations will be much faster.
+
+There are two levels of usage. The high-level one uses utility functions
+in svmutil.py and the usage is the same as the LIBSVM MATLAB interface.
+
+>>> import numpy as np
+>>> import scipy
+>>> from libsvm.svmutil import *
+# Read data in LIBSVM format
+>>> y, x = svm_read_problem('../heart_scale', return_scipy = True) # y: ndarray, x: csr_matrix
+>>> m = svm_train(y[:200], x[:200, :], '-c 4')
+>>> p_label, p_acc, p_val = svm_predict(y[200:], x[200:, :], m)
+
+# Construct problem in Scipy format
+# Dense data: numpy ndarray
+>>> y, x = np.asarray([1,-1]), np.asarray([[1,0,1], [-1,0,-1]])
+# Sparse data: scipy csr_matrix((data, (row_ind, col_ind))
+>>> y, x = np.asarray([1,-1]), scipy.sparse.csr_matrix(([1, 1, -1, -1], ([0, 0, 1, 1], [0, 2, 0, 2])))
+>>> prob  = svm_problem(y, x)
+>>> param = svm_parameter('-t 0 -c 4 -b 1')
+>>> m = svm_train(prob, param)
+
+# Precomputed kernel data (-t 4)
+# Dense data: numpy ndarray
+>>> y, x = np.asarray([1,-1]), np.asarray([[1,2,-2], [2,-2,2]])
+# Sparse data: scipy csr_matrix((data, (row_ind, col_ind))
+>>> y, x = np.asarray([1,-1]), scipy.sparse.csr_matrix(([1, 2, -2, 2, -2, 2], ([0, 0, 0, 1, 1, 1], [0, 1, 2, 0, 1, 2])))
+# isKernel=True must be set for precomputed kernel
+>>> prob  = svm_problem(y, x, isKernel=True)
+>>> param = svm_parameter('-t 4 -c 4 -b 1')
+>>> m = svm_train(prob, param)
+# For the format of precomputed kernel, please read LIBSVM README.
+
+# Apply data scaling in Scipy format
+>>> y, x = svm_read_problem('../heart_scale', return_scipy=True)
+>>> scale_param = csr_find_scale_param(x, lower=0)
+>>> scaled_x = csr_scale(x, scale_param)
+
+# Other utility functions
+>>> svm_save_model('heart_scale.model', m)
+>>> m = svm_load_model('heart_scale.model')
+>>> p_label, p_acc, p_val = svm_predict(y, x, m, '-b 1')
+>>> ACC, MSE, SCC = evaluations(y, p_label)
+
+# Getting online help
+>>> help(svm_train)
+
+The low-level use directly calls C interfaces imported by svm.py. Note that
+all arguments and return values are in ctypes format. You need to handle them
+carefully.
+
+>>> from libsvm.svm import *
+>>> prob = svm_problem(np.asarray([1,-1]), scipy.sparse.csr_matrix(([1, 1, -1, -1], ([0, 0, 1, 1], [0, 2, 0, 2]))))
+>>> param = svm_parameter('-c 4')
+>>> m = libsvm.svm_train(prob, param) # m is a ctype pointer to an svm_model
+# Convert a tuple of ndarray (index, data) to feature_nodearray, a ctypes structure
+# Note that index starts from 0, though the following example will be changed to 1:1, 3:1 internally
+>>> x0, max_idx = gen_svm_nodearray((np.asarray([0,2]), np.asarray([1,1])))
+>>> label = libsvm.svm_predict(m, x0)
+
+Design Description
+==================
+
+There are two files svm.py and svmutil.py, which respectively correspond to
+low-level and high-level use of the interface.
+
+In svm.py, we adopt the Python built-in library "ctypes," so that
+Python can directly access C structures and interface functions defined
+in svm.h.
+
+While advanced users can use structures/functions in svm.py, to
+avoid handling ctypes structures, in svmutil.py we provide some easy-to-use
+functions. The usage is similar to LIBSVM MATLAB interface.
+
+Data Structures
+===============
+
+Four data structures derived from svm.h are svm_node, svm_problem, svm_parameter,
+and svm_model. They all contain fields with the same names in svm.h. Access
+these fields carefully because you directly use a C structure instead of a
+Python object. For svm_model, accessing the field directly is not recommanded.
+Programmers should use the interface functions or methods of svm_model class
+in Python to get the values. The following description introduces additional
+fields and methods.
+
+Before using the data structures, execute the following command to load the
+LIBSVM shared library:
+
+    >>> from libsvm.svm import *
+
+- class svm_node:
+
+    Construct an svm_node.
+
+    >>> node = svm_node(idx, val)
+
+    idx: an integer indicates the feature index.
+
+    val: a float indicates the feature value.
+
+    Show the index and the value of a node.
+
+    >>> print(node)
+
+- Function: gen_svm_nodearray(xi [,feature_max=None [,isKernel=False]])
+
+    Generate a feature vector from a Python list/tuple/dictionary, numpy ndarray or tuple of (index, data):
+
+    >>> xi_ctype, max_idx = gen_svm_nodearray({1:1, 3:1, 5:-2})
+
+    xi_ctype: the returned svm_nodearray (a ctypes structure)
+
+    max_idx: the maximal feature index of xi
+
+    feature_max: if feature_max is assigned, features with indices larger than
+                 feature_max are removed.
+
+    isKernel: if isKernel == True, the list index starts from 0 for precomputed
+              kernel. Otherwise, the list index starts from 1. The default
+              value is False.
+
+- class svm_problem:
+
+    Construct an svm_problem instance
+
+    >>> prob = svm_problem(y, x)
+
+    y: a Python list/tuple/ndarray of l labels (type must be int/double).
+
+    x: 1. a list/tuple of l training instances. Feature vector of
+          each training instance is a list/tuple or dictionary.
+
+       2. an l * n numpy ndarray or scipy spmatrix (n: number of features).
+
+    Note that if your x contains sparse data (i.e., dictionary), the internal
+    ctypes data format is still sparse.
+
+    For pre-computed kernel, the isKernel flag should be set to True:
+
+    >>> prob = svm_problem(y, x, isKernel=True)
+
+    Please read LIBSVM README for more details of pre-computed kernel.
+
+- class svm_parameter:
+
+    Construct an svm_parameter instance
+
+    >>> param = svm_parameter('training_options')
+
+    If 'training_options' is empty, LIBSVM default values are applied.
+
+    Set param to LIBSVM default values.
+
+    >>> param.set_to_default_values()
+
+    Parse a string of options.
+
+    >>> param.parse_options('training_options')
+
+    Show values of parameters.
+
+    >>> print(param)
+
+- class svm_model:
+
+    There are two ways to obtain an instance of svm_model:
+
+    >>> model = svm_train(y, x)
+    >>> model = svm_load_model('model_file_name')
+
+    Note that the returned structure of interface functions
+    libsvm.svm_train and libsvm.svm_load_model is a ctypes pointer of
+    svm_model, which is different from the svm_model object returned
+    by svm_train and svm_load_model in svmutil.py. We provide a
+    function toPyModel for the conversion:
+
+    >>> model_ptr = libsvm.svm_train(prob, param)
+    >>> model = toPyModel(model_ptr)
+
+    If you obtain a model in a way other than the above approaches,
+    handle it carefully to avoid memory leak or segmentation fault.
+
+    Some interface functions to access LIBSVM models are wrapped as
+    members of the class svm_model:
+
+    >>> svm_type = model.get_svm_type()
+    >>> nr_class = model.get_nr_class()
+    >>> svr_probability = model.get_svr_probability()
+    >>> class_labels = model.get_labels()
+    >>> sv_indices = model.get_sv_indices()
+    >>> nr_sv = model.get_nr_sv()
+    >>> is_prob_model = model.is_probability_model()
+    >>> support_vector_coefficients = model.get_sv_coef()
+    >>> support_vectors = model.get_SV()
+
+Utility Functions
+=================
+
+To use utility functions, type
+
+    >>> from libsvm.svmutil import *
+
+The above command loads
+    svm_train()            : train an SVM model
+    svm_predict()          : predict testing data
+    svm_read_problem()     : read the data from a LIBSVM-format file or object.
+    svm_load_model()       : load a LIBSVM model.
+    svm_save_model()       : save model to a file.
+    evaluations()          : evaluate prediction results.
+    csr_find_scale_param() : find scaling parameter for data in csr format.
+    csr_scale()            : apply data scaling to data in csr format.
+
+- Function: svm_train
+
+    There are three ways to call svm_train()
+
+    >>> model = svm_train(y, x [, 'training_options'])
+    >>> model = svm_train(prob [, 'training_options'])
+    >>> model = svm_train(prob, param)
+
+    y: a list/tuple/ndarray of l training labels (type must be int/double).
+
+    x: 1. a list/tuple of l training instances. Feature vector of
+          each training instance is a list/tuple or dictionary.
+
+       2. an l * n numpy ndarray or scipy spmatrix (n: number of features).
+
+    training_options: a string in the same form as that for LIBSVM command
+                      mode.
+
+    prob: an svm_problem instance generated by calling
+          svm_problem(y, x).
+          For pre-computed kernel, you should use
+          svm_problem(y, x, isKernel=True)
+
+    param: an svm_parameter instance generated by calling
+           svm_parameter('training_options')
+
+    model: the returned svm_model instance. See svm.h for details of this
+           structure. If '-v' is specified, cross validation is
+           conducted and the returned model is just a scalar: cross-validation
+           accuracy for classification and mean-squared error for regression.
+
+    To train the same data many times with different
+    parameters, the second and the third ways should be faster..
+
+    Examples:
+
+    >>> y, x = svm_read_problem('../heart_scale')
+    >>> prob = svm_problem(y, x)
+    >>> param = svm_parameter('-s 3 -c 5 -h 0')
+    >>> m = svm_train(y, x, '-c 5')
+    >>> m = svm_train(prob, '-t 2 -c 5')
+    >>> m = svm_train(prob, param)
+    >>> CV_ACC = svm_train(y, x, '-v 3')
+
+- Function: svm_predict
+
+    To predict testing data with a model, use
+
+    >>> p_labs, p_acc, p_vals = svm_predict(y, x, model [,'predicting_options'])
+
+    y: a list/tuple/ndarray of l true labels (type must be int/double).
+       It is used for calculating the accuracy. Use [] if true labels are
+       unavailable.
+
+    x: 1. a list/tuple of l training instances. Feature vector of
+          each training instance is a list/tuple or dictionary.
+
+       2. an l * n numpy ndarray or scipy spmatrix (n: number of features).
+
+    predicting_options: a string of predicting options in the same format as
+                        that of LIBSVM.
+
+    model: an svm_model instance.
+
+    p_labels: a list of predicted labels
+
+    p_acc: a tuple including accuracy (for classification), mean
+           squared error, and squared correlation coefficient (for
+           regression).
+
+    p_vals: a list of decision values or probability estimates (if '-b 1'
+            is specified). If k is the number of classes in training data,
+            for decision values, each element includes results of predicting
+            k(k-1)/2 binary-class SVMs. For classification, k = 1 is a
+            special case. Decision value [+1] is returned for each testing
+            instance, instead of an empty list.
+            For probabilities, each element contains k values indicating
+            the probability that the testing instance is in each class.
+            For one-class SVM, the list has two elements indicating the
+            probabilities of normal instance/outlier.
+            Note that the order of classes is the same as the 'model.label'
+            field in the model structure.
+
+    Example:
+
+    >>> m = svm_train(y, x, '-c 5')
+    >>> p_labels, p_acc, p_vals = svm_predict(y, x, m)
+
+- Functions: svm_read_problem/svm_load_model/svm_save_model
+
+    See the usage by examples:
+
+    >>> y, x = svm_read_problem('data.txt')
+    >>> with open('data.txt') as f:
+    >>>     y, x = svm_read_problem(f)
+    >>> m = svm_load_model('model_file')
+    >>> svm_save_model('model_file', m)
+
+- Function: evaluations
+
+    Calculate some evaluations using the true values (ty) and the predicted
+    values (pv):
+
+    >>> (ACC, MSE, SCC) = evaluations(ty, pv, useScipy)
+
+    ty: a list/tuple/ndarray of true values.
+
+    pv: a list/tuple/ndarray of predicted values.
+
+    useScipy: convert ty, pv to ndarray, and use scipy functions to do the evaluation
+
+    ACC: accuracy.
+
+    MSE: mean squared error.
+
+    SCC: squared correlation coefficient.
+
+- Function: csr_find_scale_parameter/csr_scale
+
+    Scale data in csr format.
+
+    >>> param = csr_find_scale_param(x [, lower=l, upper=u])
+    >>> x = csr_scale(x, param)
+
+    x: a csr_matrix of data.
+
+    l: x scaling lower limit; default -1.
+
+    u: x scaling upper limit; default 1.
+
+    The scaling process is: x * diag(coef) + ones(l, 1) * offset'
+
+    param: a dictionary of scaling parameters, where param['coef'] = coef and param['offset'] = offset.
+
+    coef: a scipy array of scaling coefficients.
+
+    offset: a scipy array of scaling offsets.
+
+Additional Information
+======================
+
+This interface was originally written by Hsiang-Fu Yu from Department of Computer
+Science, National Taiwan University. If you find this tool useful, please
+cite LIBSVM as follows
+
+Chih-Chung Chang and Chih-Jen Lin, LIBSVM : a library for support
+vector machines. ACM Transactions on Intelligent Systems and
+Technology, 2:27:1--27:27, 2011. Software available at
+http://www.csie.ntu.edu.tw/~cjlin/libsvm
+
+For any question, please contact Chih-Jen Lin <cjlin@csie.ntu.edu.tw>,
+or check the FAQ page:
+
+http://www.csie.ntu.edu.tw/~cjlin/libsvm/faq.html

libsvm-3.36/python/libsvm/commonutil.py

@@ -0,0 +1,189 @@
+from __future__ import print_function
+from array import array
+import sys
+
+try:
+    import numpy as np
+    import scipy
+    from scipy import sparse
+except:
+    scipy = None
+
+
+__all__ = ['svm_read_problem', 'evaluations', 'csr_find_scale_param', 'csr_scale']
+
+def svm_read_problem(data_source, return_scipy=False):
+    """
+    svm_read_problem(data_source, return_scipy=False) -> [y, x], y: list, x: list of dictionary
+    svm_read_problem(data_source, return_scipy=True)  -> [y, x], y: ndarray, x: csr_matrix
+
+    Read LIBSVM-format data from data_source and return labels y
+    and data instances x.
+    """
+    if scipy != None and return_scipy:
+        prob_y = array('d')
+        prob_x = array('d')
+        row_ptr = array('l', [0])
+        col_idx = array('l')
+    else:
+        prob_y = []
+        prob_x = []
+        row_ptr = [0]
+        col_idx = []
+    indx_start = 1
+
+    if hasattr(data_source, "read"):
+        file = data_source
+    else:
+        file = open(data_source)
+    try:
+        for line in file:
+            line = line.split(None, 1)
+            # In case an instance with all zero features
+            if len(line) == 1: line += ['']
+            label, features = line
+            prob_y.append(float(label))
+            if scipy != None and return_scipy:
+                nz = 0
+                for e in features.split():
+                    ind, val = e.split(":")
+                    if ind == '0':
+                        indx_start = 0
+                    val = float(val)
+                    if val != 0:
+                        col_idx.append(int(ind)-indx_start)
+                        prob_x.append(val)
+                        nz += 1
+                row_ptr.append(row_ptr[-1]+nz)
+            else:
+                xi = {}
+                for e in features.split():
+                    ind, val = e.split(":")
+                    xi[int(ind)] = float(val)
+                prob_x += [xi]
+    except Exception as err_msg:
+        raise err_msg
+    finally:
+        if not hasattr(data_source, "read"):
+            # close file only if it was created by us
+            file.close()
+
+    if scipy != None and return_scipy:
+        prob_y = np.frombuffer(prob_y, dtype='d')
+        prob_x = np.frombuffer(prob_x, dtype='d')
+        col_idx = np.frombuffer(col_idx, dtype='l')
+        row_ptr = np.frombuffer(row_ptr, dtype='l')
+        prob_x = sparse.csr_matrix((prob_x, col_idx, row_ptr))
+    return (prob_y, prob_x)
+
+def evaluations_scipy(ty, pv):
+    """
+    evaluations_scipy(ty, pv) -> (ACC, MSE, SCC)
+    ty, pv: ndarray
+
+    Calculate accuracy, mean squared error and squared correlation coefficient
+    using the true values (ty) and predicted values (pv).
+    """
+    if not (scipy != None and isinstance(ty, np.ndarray) and isinstance(pv, np.ndarray)):
+        raise TypeError("type of ty and pv must be ndarray")
+    if len(ty) != len(pv):
+        raise ValueError("len(ty) must be equal to len(pv)")
+    ACC = 100.0*(ty == pv).mean()
+    MSE = ((ty - pv)**2).mean()
+    l = len(ty)
+    sumv = pv.sum()
+    sumy = ty.sum()
+    sumvy = (pv*ty).sum()
+    sumvv = (pv*pv).sum()
+    sumyy = (ty*ty).sum()
+    with np.errstate(all = 'raise'):
+        try:
+            SCC = ((l*sumvy-sumv*sumy)*(l*sumvy-sumv*sumy))/((l*sumvv-sumv*sumv)*(l*sumyy-sumy*sumy))
+        except:
+            SCC = float('nan')
+    return (float(ACC), float(MSE), float(SCC))
+
+def evaluations(ty, pv, useScipy = True):
+    """
+    evaluations(ty, pv, useScipy) -> (ACC, MSE, SCC)
+    ty, pv: list, tuple or ndarray
+    useScipy: convert ty, pv to ndarray, and use scipy functions for the evaluation
+
+    Calculate accuracy, mean squared error and squared correlation coefficient
+    using the true values (ty) and predicted values (pv).
+    """
+    if scipy != None and useScipy:
+        return evaluations_scipy(np.asarray(ty), np.asarray(pv))
+    if len(ty) != len(pv):
+        raise ValueError("len(ty) must be equal to len(pv)")
+    total_correct = total_error = 0
+    sumv = sumy = sumvv = sumyy = sumvy = 0
+    for v, y in zip(pv, ty):
+        if y == v:
+            total_correct += 1
+        total_error += (v-y)*(v-y)
+        sumv += v
+        sumy += y
+        sumvv += v*v
+        sumyy += y*y
+        sumvy += v*y
+    l = len(ty)
+    ACC = 100.0*total_correct/l
+    MSE = total_error/l
+    try:
+        SCC = ((l*sumvy-sumv*sumy)*(l*sumvy-sumv*sumy))/((l*sumvv-sumv*sumv)*(l*sumyy-sumy*sumy))
+    except:
+        SCC = float('nan')
+    return (float(ACC), float(MSE), float(SCC))
+
+def csr_find_scale_param(x, lower=-1, upper=1):
+    assert isinstance(x, sparse.csr_matrix)
+    assert lower < upper
+    l, n = x.shape
+    feat_min = x.min(axis=0).toarray().flatten()
+    feat_max = x.max(axis=0).toarray().flatten()
+    coef = (feat_max - feat_min) / (upper - lower)
+    coef[coef != 0] = 1.0 / coef[coef != 0]
+
+    # (x - ones(l,1) * feat_min') * diag(coef) + lower
+    # = x * diag(coef) - ones(l, 1) * (feat_min' * diag(coef)) + lower
+    # = x * diag(coef) + ones(l, 1) * (-feat_min' * diag(coef) + lower)
+    # = x * diag(coef) + ones(l, 1) * offset'
+    offset = -feat_min * coef + lower
+    offset[coef == 0] = 0
+
+    if sum(offset != 0) * l > 3 * x.getnnz():
+        print(
+            "WARNING: The #nonzeros of the scaled data is at least 2 times larger than the original one.\n"
+            "If feature values are non-negative and sparse, set lower=0 rather than the default lower=-1.",
+            file=sys.stderr)
+
+    return {'coef':coef, 'offset':offset}
+
+def csr_scale(x, scale_param):
+    assert isinstance(x, sparse.csr_matrix)
+
+    offset = scale_param['offset']
+    coef = scale_param['coef']
+    assert len(coef) == len(offset)
+
+    l, n = x.shape
+
+    if not n == len(coef):
+        print("WARNING: The dimension of scaling parameters and feature number do not match.", file=sys.stderr)
+        coef = coef.resize(n) # zeros padded if n > len(coef)
+        offset = offset.resize(n)
+
+    # scaled_x = x * diag(coef) + ones(l, 1) * offset'
+    offset = sparse.csr_matrix(offset.reshape(1, n))
+    offset = sparse.vstack([offset] * l, format='csr', dtype=x.dtype)
+    scaled_x = x.dot(sparse.diags(coef, 0, shape=(n, n))) + offset
+
+    if scaled_x.getnnz() > x.getnnz():
+        print(
+            "WARNING: original #nonzeros %d\n" % x.getnnz() +
+            "       > new      #nonzeros %d\n" % scaled_x.getnnz() +
+            "If feature values are non-negative and sparse, get scale_param by setting lower=0 rather than the default lower=-1.",
+            file=sys.stderr)
+
+    return scaled_x

libsvm-3.36/python/libsvm/svm.py

@@ -0,0 +1,465 @@
+from ctypes import *
+from ctypes.util import find_library
+from os import path
+from glob import glob
+from enum import IntEnum
+import sys
+
+try:
+    import numpy as np
+    import scipy
+    from scipy import sparse
+except:
+    scipy = None
+
+
+if sys.version_info[0] < 3:
+    range = xrange
+    from itertools import izip as zip
+
+__all__ = ['libsvm', 'svm_problem', 'svm_parameter',
+           'toPyModel', 'gen_svm_nodearray', 'print_null', 'svm_node', 'svm_forms',
+            'PRINT_STRING_FUN', 'kernel_names', 'c_double', 'svm_model']
+
+try:
+    dirname = path.dirname(path.abspath(__file__))
+    dynamic_lib_name = 'clib.cp*'
+    path_to_so = glob(path.join(dirname, dynamic_lib_name))[0]
+    libsvm = CDLL(path_to_so)
+except:
+    try:
+        if sys.platform == 'win32':
+            libsvm = CDLL(path.join(dirname, r'..\..\windows\libsvm.dll'))
+        else:
+            libsvm = CDLL(path.join(dirname, '../../libsvm.so.4'))
+    except:
+    # For unix the prefix 'lib' is not considered.
+        if find_library('svm'):
+            libsvm = CDLL(find_library('svm'))
+        elif find_library('libsvm'):
+            libsvm = CDLL(find_library('libsvm'))
+        else:
+            raise Exception('LIBSVM library not found.')
+
+class svm_forms(IntEnum):
+    C_SVC = 0
+    NU_SVC = 1
+    ONE_CLASS = 2
+    EPSILON_SVR = 3
+    NU_SVR = 4
+
+class kernel_names(IntEnum):
+    LINEAR = 0
+    POLY = 1
+    RBF = 2
+    SIGMOID = 3
+    PRECOMPUTED = 4
+
+PRINT_STRING_FUN = CFUNCTYPE(None, c_char_p)
+def print_null(s):
+    return
+
+# In multi-threading, all threads share the same memory space of
+# the dynamic library (libsvm). Thus, we use a module-level
+# variable to keep a reference to ctypes print_null, preventing
+# python from garbage collecting it in thread B while thread A
+# still needs it. Check the usage of svm_set_print_string_function()
+# in LIBSVM README for details.
+ctypes_print_null = PRINT_STRING_FUN(print_null)
+
+def genFields(names, types):
+    return list(zip(names, types))
+
+def fillprototype(f, restype, argtypes):
+    f.restype = restype
+    f.argtypes = argtypes
+
+class svm_node(Structure):
+    _names = ["index", "value"]
+    _types = [c_int, c_double]
+    _fields_ = genFields(_names, _types)
+
+    def __init__(self, index=-1, value=0):
+        self.index, self.value = index, value
+
+    def __str__(self):
+        return '%d:%g' % (self.index, self.value)
+
+def gen_svm_nodearray(xi, feature_max=None, isKernel=False):
+    if feature_max:
+        assert(isinstance(feature_max, int))
+
+    xi_shift = 0 # ensure correct indices of xi
+    if scipy and isinstance(xi, tuple) and len(xi) == 2\
+            and isinstance(xi[0], np.ndarray) and isinstance(xi[1], np.ndarray): # for a sparse vector
+        if not isKernel:
+            index_range = xi[0] + 1 # index starts from 1
+        else:
+            index_range = xi[0] # index starts from 0 for precomputed kernel
+        if feature_max:
+            index_range = index_range[np.where(index_range <= feature_max)]
+    elif scipy and isinstance(xi, np.ndarray):
+        if not isKernel:
+            xi_shift = 1
+            index_range = xi.nonzero()[0] + 1 # index starts from 1
+        else:
+            index_range = np.arange(0, len(xi)) # index starts from 0 for precomputed kernel
+        if feature_max:
+            index_range = index_range[np.where(index_range <= feature_max)]
+    elif isinstance(xi, (dict, list, tuple)):
+        if isinstance(xi, dict):
+            index_range = sorted(xi.keys())
+        elif isinstance(xi, (list, tuple)):
+            if not isKernel:
+                xi_shift = 1
+                index_range = range(1, len(xi) + 1) # index starts from 1
+            else:
+                index_range = range(0, len(xi)) # index starts from 0 for precomputed kernel
+
+        if feature_max:
+            index_range = list(filter(lambda j: j <= feature_max, index_range))
+        if not isKernel:
+            index_range = list(filter(lambda j:xi[j-xi_shift] != 0, index_range))
+    else:
+        raise TypeError('xi should be a dictionary, list, tuple, 1-d numpy array, or tuple of (index, data)')
+
+    ret = (svm_node*(len(index_range)+1))()
+    ret[-1].index = -1
+
+    if scipy and isinstance(xi, tuple) and len(xi) == 2\
+            and isinstance(xi[0], np.ndarray) and isinstance(xi[1], np.ndarray): # for a sparse vector
+        # since xi=(indices, values), we must sort them simultaneously.
+        for idx, arg in enumerate(np.argsort(index_range)):
+            ret[idx].index = index_range[arg]
+            ret[idx].value = (xi[1])[arg]
+    else:
+        for idx, j in enumerate(index_range):
+            ret[idx].index = j
+            ret[idx].value = xi[j - xi_shift]
+
+    max_idx = 0
+    if len(index_range) > 0:
+        max_idx = index_range[-1]
+    return ret, max_idx
+
+try:
+    from numba import jit
+    jit_enabled = True
+except:
+    # We need to support two cases: when jit is called with no arguments, and when jit is called with
+    # a keyword argument.
+    def jit(func=None, *args, **kwargs):
+        if func is None:
+            # This handles the case where jit is used with parentheses: @jit(nopython=True)
+            return lambda x: x
+        else:
+            # This handles the case where jit is used without parentheses: @jit
+            return func
+    jit_enabled = False
+
+@jit(nopython=True)
+def csr_to_problem_jit(l, x_val, x_ind, x_rowptr, prob_val, prob_ind, prob_rowptr, indx_start):
+    for i in range(l):
+        b1,e1 = x_rowptr[i], x_rowptr[i+1]
+        b2,e2 = prob_rowptr[i], prob_rowptr[i+1]-1
+        for j in range(b1,e1):
+            prob_ind[j-b1+b2] = x_ind[j]+indx_start
+            prob_val[j-b1+b2] = x_val[j]
+def csr_to_problem_nojit(l, x_val, x_ind, x_rowptr, prob_val, prob_ind, prob_rowptr, indx_start):
+    for i in range(l):
+        x_slice = slice(x_rowptr[i], x_rowptr[i+1])
+        prob_slice = slice(prob_rowptr[i], prob_rowptr[i+1]-1)
+        prob_ind[prob_slice] = x_ind[x_slice]+indx_start
+        prob_val[prob_slice] = x_val[x_slice]
+
+def csr_to_problem(x, prob, isKernel):
+    if not x.has_sorted_indices:
+        x.sort_indices()
+
+    # Extra space for termination node and (possibly) bias term
+    x_space = prob.x_space = np.empty((x.nnz+x.shape[0]), dtype=svm_node)
+    # rowptr has to be a 64bit integer because it will later be used for pointer arithmetic,
+    # which overflows when the added pointer points to an address that is numerically high.
+    prob.rowptr = x.indptr.astype(np.int64, copy=True)
+    prob.rowptr[1:] += np.arange(1,x.shape[0]+1)
+    prob_ind = x_space["index"]
+    prob_val = x_space["value"]
+    prob_ind[:] = -1
+    if not isKernel:
+        indx_start = 1 # index starts from 1
+    else:
+        indx_start = 0 # index starts from 0 for precomputed kernel
+    if jit_enabled:
+        csr_to_problem_jit(x.shape[0], x.data, x.indices, x.indptr, prob_val, prob_ind, prob.rowptr, indx_start)
+    else:
+        csr_to_problem_nojit(x.shape[0], x.data, x.indices, x.indptr, prob_val, prob_ind, prob.rowptr, indx_start)
+
+class svm_problem(Structure):
+    _names = ["l", "y", "x"]
+    _types = [c_int, POINTER(c_double), POINTER(POINTER(svm_node))]
+    _fields_ = genFields(_names, _types)
+
+    def __init__(self, y, x, isKernel=False):
+        if (not isinstance(y, (list, tuple))) and (not (scipy and isinstance(y, np.ndarray))):
+            raise TypeError("type of y: {0} is not supported!".format(type(y)))
+
+        if isinstance(x, (list, tuple)):
+            if len(y) != len(x):
+                raise ValueError("len(y) != len(x)")
+        elif scipy != None and isinstance(x, (np.ndarray, sparse.spmatrix)):
+            if len(y) != x.shape[0]:
+                raise ValueError("len(y) != len(x)")
+            if isinstance(x, np.ndarray):
+                x = np.ascontiguousarray(x) # enforce row-major
+            if isinstance(x, sparse.spmatrix):
+                x = x.tocsr()
+                pass
+        else:
+            raise TypeError("type of x: {0} is not supported!".format(type(x)))
+        self.l = l = len(y)
+
+        max_idx = 0
+        x_space = self.x_space = []
+        if scipy != None and isinstance(x, sparse.csr_matrix):
+            csr_to_problem(x, self, isKernel)
+            max_idx = x.shape[1]
+        else:
+            for i, xi in enumerate(x):
+                tmp_xi, tmp_idx = gen_svm_nodearray(xi,isKernel=isKernel)
+                x_space += [tmp_xi]
+                max_idx = max(max_idx, tmp_idx)
+        self.n = max_idx
+
+        self.y = (c_double * l)()
+        if scipy != None and isinstance(y, np.ndarray):
+            np.ctypeslib.as_array(self.y, (self.l,))[:] = y
+        else:
+            for i, yi in enumerate(y): self.y[i] = yi
+
+        self.x = (POINTER(svm_node) * l)()
+        if scipy != None and isinstance(x, sparse.csr_matrix):
+            base = addressof(self.x_space.ctypes.data_as(POINTER(svm_node))[0])
+            x_ptr = cast(self.x, POINTER(c_uint64))
+            x_ptr = np.ctypeslib.as_array(x_ptr,(self.l,))
+            x_ptr[:] = self.rowptr[:-1]*sizeof(svm_node)+base
+        else:
+            for i, xi in enumerate(self.x_space): self.x[i] = xi
+
+class svm_parameter(Structure):
+    _names = ["svm_type", "kernel_type", "degree", "gamma", "coef0",
+            "cache_size", "eps", "C", "nr_weight", "weight_label", "weight",
+            "nu", "p", "shrinking", "probability"]
+    _types = [c_int, c_int, c_int, c_double, c_double,
+            c_double, c_double, c_double, c_int, POINTER(c_int), POINTER(c_double),
+            c_double, c_double, c_int, c_int]
+    _fields_ = genFields(_names, _types)
+
+    def __init__(self, options = None):
+        if options == None:
+            options = ''
+        self.parse_options(options)
+
+    def __str__(self):
+        s = ''
+        attrs = svm_parameter._names + list(self.__dict__.keys())
+        values = map(lambda attr: getattr(self, attr), attrs)
+        for attr, val in zip(attrs, values):
+            s += (' %s: %s\n' % (attr, val))
+        s = s.strip()
+
+        return s
+
+    def set_to_default_values(self):
+        self.svm_type = svm_forms.C_SVC;
+        self.kernel_type = kernel_names.RBF
+        self.degree = 3
+        self.gamma = 0
+        self.coef0 = 0
+        self.nu = 0.5
+        self.cache_size = 100
+        self.C = 1
+        self.eps = 0.001
+        self.p = 0.1
+        self.shrinking = 1
+        self.probability = 0
+        self.nr_weight = 0
+        self.weight_label = None
+        self.weight = None
+        self.cross_validation = False
+        self.nr_fold = 0
+        self.print_func = cast(None, PRINT_STRING_FUN)
+
+    def parse_options(self, options):
+        if isinstance(options, list):
+            argv = options
+        elif isinstance(options, str):
+            argv = options.split()
+        else:
+            raise TypeError("arg 1 should be a list or a str.")
+        self.set_to_default_values()
+        self.print_func = cast(None, PRINT_STRING_FUN)
+        weight_label = []
+        weight = []
+
+        i = 0
+        while i < len(argv):
+            if argv[i] == "-s":
+                i = i + 1
+                self.svm_type = svm_forms(int(argv[i]))
+            elif argv[i] == "-t":
+                i = i + 1
+                self.kernel_type = kernel_names(int(argv[i]))
+            elif argv[i] == "-d":
+                i = i + 1
+                self.degree = int(argv[i])
+            elif argv[i] == "-g":
+                i = i + 1
+                self.gamma = float(argv[i])
+            elif argv[i] == "-r":
+                i = i + 1
+                self.coef0 = float(argv[i])
+            elif argv[i] == "-n":
+                i = i + 1
+                self.nu = float(argv[i])
+            elif argv[i] == "-m":
+                i = i + 1
+                self.cache_size = float(argv[i])
+            elif argv[i] == "-c":
+                i = i + 1
+                self.C = float(argv[i])
+            elif argv[i] == "-e":
+                i = i + 1
+                self.eps = float(argv[i])
+            elif argv[i] == "-p":
+                i = i + 1
+                self.p = float(argv[i])
+            elif argv[i] == "-h":
+                i = i + 1
+                self.shrinking = int(argv[i])
+            elif argv[i] == "-b":
+                i = i + 1
+                self.probability = int(argv[i])
+            elif argv[i] == "-q":
+                self.print_func = ctypes_print_null
+            elif argv[i] == "-v":
+                i = i + 1
+                self.cross_validation = 1
+                self.nr_fold = int(argv[i])
+                if self.nr_fold < 2:
+                    raise ValueError("n-fold cross validation: n must >= 2")
+            elif argv[i].startswith("-w"):
+                i = i + 1
+                self.nr_weight += 1
+                weight_label += [int(argv[i-1][2:])]
+                weight += [float(argv[i])]
+            else:
+                raise ValueError("Wrong options")
+            i += 1
+
+        libsvm.svm_set_print_string_function(self.print_func)
+        self.weight_label = (c_int*self.nr_weight)()
+        self.weight = (c_double*self.nr_weight)()
+        for i in range(self.nr_weight):
+            self.weight[i] = weight[i]
+            self.weight_label[i] = weight_label[i]
+
+class svm_model(Structure):
+    _names = ['param', 'nr_class', 'l', 'SV', 'sv_coef', 'rho',
+            'probA', 'probB', 'prob_density_marks', 'sv_indices',
+            'label', 'nSV', 'free_sv']
+    _types = [svm_parameter, c_int, c_int, POINTER(POINTER(svm_node)),
+            POINTER(POINTER(c_double)), POINTER(c_double),
+            POINTER(c_double), POINTER(c_double), POINTER(c_double),
+            POINTER(c_int), POINTER(c_int), POINTER(c_int), c_int]
+    _fields_ = genFields(_names, _types)
+
+    def __init__(self):
+        self.__createfrom__ = 'python'
+
+    def __del__(self):
+        # free memory created by C to avoid memory leak
+        if hasattr(self, '__createfrom__') and self.__createfrom__ == 'C':
+            libsvm.svm_free_and_destroy_model(pointer(pointer(self)))
+
+    def get_svm_type(self):
+        return libsvm.svm_get_svm_type(self)
+
+    def get_nr_class(self):
+        return libsvm.svm_get_nr_class(self)
+
+    def get_svr_probability(self):
+        return libsvm.svm_get_svr_probability(self)
+
+    def get_labels(self):
+        nr_class = self.get_nr_class()
+        labels = (c_int * nr_class)()
+        libsvm.svm_get_labels(self, labels)
+        return labels[:nr_class]
+
+    def get_sv_indices(self):
+        total_sv = self.get_nr_sv()
+        sv_indices = (c_int * total_sv)()
+        libsvm.svm_get_sv_indices(self, sv_indices)
+        return sv_indices[:total_sv]
+
+    def get_nr_sv(self):
+        return libsvm.svm_get_nr_sv(self)
+
+    def is_probability_model(self):
+        return (libsvm.svm_check_probability_model(self) == 1)
+
+    def get_sv_coef(self):
+        return [tuple(self.sv_coef[j][i] for j in range(self.nr_class - 1))
+                for i in range(self.l)]
+
+    def get_SV(self):
+        result = []
+        for sparse_sv in self.SV[:self.l]:
+            row = dict()
+
+            i = 0
+            while True:
+                if sparse_sv[i].index == -1:
+                    break
+                row[sparse_sv[i].index] = sparse_sv[i].value
+                i += 1
+
+            result.append(row)
+        return result
+
+def toPyModel(model_ptr):
+    """
+    toPyModel(model_ptr) -> svm_model
+
+    Convert a ctypes POINTER(svm_model) to a Python svm_model
+    """
+    if bool(model_ptr) == False:
+        raise ValueError("Null pointer")
+    m = model_ptr.contents
+    m.__createfrom__ = 'C'
+    return m
+
+fillprototype(libsvm.svm_train, POINTER(svm_model), [POINTER(svm_problem), POINTER(svm_parameter)])
+fillprototype(libsvm.svm_cross_validation, None, [POINTER(svm_problem), POINTER(svm_parameter), c_int, POINTER(c_double)])
+
+fillprototype(libsvm.svm_save_model, c_int, [c_char_p, POINTER(svm_model)])
+fillprototype(libsvm.svm_load_model, POINTER(svm_model), [c_char_p])
+
+fillprototype(libsvm.svm_get_svm_type, c_int, [POINTER(svm_model)])
+fillprototype(libsvm.svm_get_nr_class, c_int, [POINTER(svm_model)])
+fillprototype(libsvm.svm_get_labels, None, [POINTER(svm_model), POINTER(c_int)])
+fillprototype(libsvm.svm_get_sv_indices, None, [POINTER(svm_model), POINTER(c_int)])
+fillprototype(libsvm.svm_get_nr_sv, c_int, [POINTER(svm_model)])
+fillprototype(libsvm.svm_get_svr_probability, c_double, [POINTER(svm_model)])
+
+fillprototype(libsvm.svm_predict_values, c_double, [POINTER(svm_model), POINTER(svm_node), POINTER(c_double)])
+fillprototype(libsvm.svm_predict, c_double, [POINTER(svm_model), POINTER(svm_node)])
+fillprototype(libsvm.svm_predict_probability, c_double, [POINTER(svm_model), POINTER(svm_node), POINTER(c_double)])
+
+fillprototype(libsvm.svm_free_model_content, None, [POINTER(svm_model)])
+fillprototype(libsvm.svm_free_and_destroy_model, None, [POINTER(POINTER(svm_model))])
+fillprototype(libsvm.svm_destroy_param, None, [POINTER(svm_parameter)])
+
+fillprototype(libsvm.svm_check_parameter, c_char_p, [POINTER(svm_problem), POINTER(svm_parameter)])
+fillprototype(libsvm.svm_check_probability_model, c_int, [POINTER(svm_model)])
+fillprototype(libsvm.svm_set_print_string_function, None, [PRINT_STRING_FUN])

libsvm-3.36/python/libsvm/svmutil.py

@@ -0,0 +1,263 @@
+import os, sys
+from .svm import *
+from .svm import __all__ as svm_all
+from .commonutil import *
+from .commonutil import __all__ as common_all
+
+try:
+    import numpy as np
+    import scipy
+    from scipy import sparse
+except:
+    scipy = None
+
+
+if sys.version_info[0] < 3:
+    range = xrange
+    from itertools import izip as zip
+    _cstr = lambda s: s.encode("utf-8") if isinstance(s,unicode) else str(s)
+else:
+    _cstr = lambda s: bytes(s, "utf-8")
+
+__all__ = ['svm_load_model', 'svm_predict', 'svm_save_model', 'svm_train'] + svm_all + common_all
+
+
+def svm_load_model(model_file_name):
+    """
+    svm_load_model(model_file_name) -> model
+
+    Load a LIBSVM model from model_file_name and return.
+    """
+    model = libsvm.svm_load_model(_cstr(model_file_name))
+    if not model:
+        print("can't open model file %s" % model_file_name)
+        return None
+    model = toPyModel(model)
+    return model
+
+def svm_save_model(model_file_name, model):
+    """
+    svm_save_model(model_file_name, model) -> None
+
+    Save a LIBSVM model to the file model_file_name.
+    """
+    libsvm.svm_save_model(_cstr(model_file_name), model)
+
+def svm_train(arg1, arg2=None, arg3=None):
+    """
+    svm_train(y, x [, options]) -> model | ACC | MSE
+
+    y: a list/tuple/ndarray of l true labels (type must be int/double).
+
+    x: 1. a list/tuple of l training instances. Feature vector of
+          each training instance is a list/tuple or dictionary.
+
+       2. an l * n numpy ndarray or scipy spmatrix (n: number of features).
+
+    svm_train(prob [, options]) -> model | ACC | MSE
+    svm_train(prob, param) -> model | ACC| MSE
+
+    Train an SVM model from data (y, x) or an svm_problem prob using
+    'options' or an svm_parameter param.
+    If '-v' is specified in 'options' (i.e., cross validation)
+    either accuracy (ACC) or mean-squared error (MSE) is returned.
+    options:
+        -s svm_type : set type of SVM (default 0)
+            0 -- C-SVC        (multi-class classification)
+            1 -- nu-SVC        (multi-class classification)
+            2 -- one-class SVM
+            3 -- epsilon-SVR    (regression)
+            4 -- nu-SVR        (regression)
+        -t kernel_type : set type of kernel function (default 2)
+            0 -- linear: u'*v
+            1 -- polynomial: (gamma*u'*v + coef0)^degree
+            2 -- radial basis function: exp(-gamma*|u-v|^2)
+            3 -- sigmoid: tanh(gamma*u'*v + coef0)
+            4 -- precomputed kernel (kernel values in training_set_file)
+        -d degree : set degree in kernel function (default 3)
+        -g gamma : set gamma in kernel function (default 1/num_features)
+        -r coef0 : set coef0 in kernel function (default 0)
+        -c cost : set the parameter C of C-SVC, epsilon-SVR, and nu-SVR (default 1)
+        -n nu : set the parameter nu of nu-SVC, one-class SVM, and nu-SVR (default 0.5)
+        -p epsilon : set the epsilon in loss function of epsilon-SVR (default 0.1)
+        -m cachesize : set cache memory size in MB (default 100)
+        -e epsilon : set tolerance of termination criterion (default 0.001)
+        -h shrinking : whether to use the shrinking heuristics, 0 or 1 (default 1)
+        -b probability_estimates : whether to train a model for probability estimates, 0 or 1 (default 0)
+        -wi weight : set the parameter C of class i to weight*C, for C-SVC (default 1)
+        -v n: n-fold cross validation mode
+        -q : quiet mode (no outputs)
+    """
+    prob, param = None, None
+    if isinstance(arg1, (list, tuple)) or (scipy and isinstance(arg1, np.ndarray)):
+        assert isinstance(arg2, (list, tuple)) or (scipy and isinstance(arg2, (np.ndarray, sparse.spmatrix)))
+        y, x, options = arg1, arg2, arg3
+        param = svm_parameter(options)
+        prob = svm_problem(y, x, isKernel=(param.kernel_type == kernel_names.PRECOMPUTED))
+    elif isinstance(arg1, svm_problem):
+        prob = arg1
+        if isinstance(arg2, svm_parameter):
+            param = arg2
+        else:
+            param = svm_parameter(arg2)
+    if prob == None or param == None:
+        raise TypeError("Wrong types for the arguments")
+
+    if param.kernel_type == kernel_names.PRECOMPUTED:
+        for i in range(prob.l):
+            xi = prob.x[i]
+            idx, val = xi[0].index, xi[0].value
+            if idx != 0:
+                raise ValueError('Wrong input format: first column must be 0:sample_serial_number')
+            if val <= 0 or val > prob.n:
+                raise ValueError('Wrong input format: sample_serial_number out of range')
+
+    if param.gamma == 0 and prob.n > 0:
+        param.gamma = 1.0 / prob.n
+    libsvm.svm_set_print_string_function(param.print_func)
+    err_msg = libsvm.svm_check_parameter(prob, param)
+    if err_msg:
+        raise ValueError('Error: %s' % err_msg)
+
+    if param.cross_validation:
+        l, nr_fold = prob.l, param.nr_fold
+        target = (c_double * l)()
+        libsvm.svm_cross_validation(prob, param, nr_fold, target)
+        ACC, MSE, SCC = evaluations(prob.y[:l], target[:l])
+        if param.svm_type in [svm_forms.EPSILON_SVR, svm_forms.NU_SVR]:
+            print("Cross Validation Mean squared error = %g" % MSE)
+            print("Cross Validation Squared correlation coefficient = %g" % SCC)
+            return MSE
+        else:
+            print("Cross Validation Accuracy = %g%%" % ACC)
+            return ACC
+    else:
+        m = libsvm.svm_train(prob, param)
+        m = toPyModel(m)
+
+        # If prob is destroyed, data including SVs pointed by m can remain.
+        m.x_space = prob.x_space
+        return m
+
+def svm_predict(y, x, m, options=""):
+    """
+    svm_predict(y, x, m [, options]) -> (p_labels, p_acc, p_vals)
+
+    y: a list/tuple/ndarray of l true labels (type must be int/double).
+       It is used for calculating the accuracy. Use [] if true labels are
+       unavailable.
+
+    x: 1. a list/tuple of l training instances. Feature vector of
+          each training instance is a list/tuple or dictionary.
+
+       2. an l * n numpy ndarray or scipy spmatrix (n: number of features).
+
+    Predict data (y, x) with the SVM model m.
+    options:
+        -b probability_estimates: whether to predict probability estimates,
+            0 or 1 (default 0).
+        -q : quiet mode (no outputs).
+
+    The return tuple contains
+    p_labels: a list of predicted labels
+    p_acc: a tuple including  accuracy (for classification), mean-squared
+           error, and squared correlation coefficient (for regression).
+    p_vals: a list of decision values or probability estimates (if '-b 1'
+            is specified). If k is the number of classes, for decision values,
+            each element includes results of predicting k(k-1)/2 binary-class
+            SVMs. For probabilities, each element contains k values indicating
+            the probability that the testing instance is in each class.
+            Note that the order of classes here is the same as 'model.label'
+            field in the model structure.
+    """
+
+    def info(s):
+        print(s)
+
+    if scipy and isinstance(x, np.ndarray):
+        x = np.ascontiguousarray(x) # enforce row-major
+    elif sparse and isinstance(x, sparse.spmatrix):
+        x = x.tocsr()
+    elif not isinstance(x, (list, tuple)):
+        raise TypeError("type of x: {0} is not supported!".format(type(x)))
+
+    if (not isinstance(y, (list, tuple))) and (not (scipy and isinstance(y, np.ndarray))):
+        raise TypeError("type of y: {0} is not supported!".format(type(y)))
+
+    predict_probability = 0
+    argv = options.split()
+    i = 0
+    while i < len(argv):
+        if argv[i] == '-b':
+            i += 1
+            predict_probability = int(argv[i])
+        elif argv[i] == '-q':
+            info = print_null
+        else:
+            raise ValueError("Wrong options")
+        i+=1
+
+    svm_type = m.get_svm_type()
+    is_prob_model = m.is_probability_model()
+    nr_class = m.get_nr_class()
+    pred_labels = []
+    pred_values = []
+
+    if scipy and isinstance(x, sparse.spmatrix):
+        nr_instance = x.shape[0]
+    else:
+        nr_instance = len(x)
+
+    if predict_probability:
+        if not is_prob_model:
+            raise ValueError("Model does not support probabiliy estimates")
+
+        if svm_type in [svm_forms.NU_SVR, svm_forms.EPSILON_SVR]:
+            info("Prob. model for test data: target value = predicted value + z,\n"
+            "z: Laplace distribution e^(-|z|/sigma)/(2sigma),sigma=%g" % m.get_svr_probability());
+            nr_class = 0
+
+        prob_estimates = (c_double * nr_class)()
+        for i in range(nr_instance):
+            if scipy and isinstance(x, sparse.spmatrix):
+                indslice = slice(x.indptr[i], x.indptr[i+1])
+                xi, idx = gen_svm_nodearray((x.indices[indslice], x.data[indslice]), isKernel=(m.param.kernel_type == kernel_names.PRECOMPUTED))
+            else:
+                xi, idx = gen_svm_nodearray(x[i], isKernel=(m.param.kernel_type == kernel_names.PRECOMPUTED))
+            label = libsvm.svm_predict_probability(m, xi, prob_estimates)
+            values = prob_estimates[:nr_class]
+            pred_labels += [label]
+            pred_values += [values]
+    else:
+        if is_prob_model:
+            info("Model supports probability estimates, but disabled in predicton.")
+        if svm_type in [svm_forms.ONE_CLASS, svm_forms.EPSILON_SVR, svm_forms.NU_SVC]:
+            nr_classifier = 1
+        else:
+            nr_classifier = nr_class*(nr_class-1)//2
+        dec_values = (c_double * nr_classifier)()
+        for i in range(nr_instance):
+            if scipy and isinstance(x, sparse.spmatrix):
+                indslice = slice(x.indptr[i], x.indptr[i+1])
+                xi, idx = gen_svm_nodearray((x.indices[indslice], x.data[indslice]), isKernel=(m.param.kernel_type == kernel_names.PRECOMPUTED))
+            else:
+                xi, idx = gen_svm_nodearray(x[i], isKernel=(m.param.kernel_type == kernel_names.PRECOMPUTED))
+            label = libsvm.svm_predict_values(m, xi, dec_values)
+            if(nr_class == 1):
+                values = [1]
+            else:
+                values = dec_values[:nr_classifier]
+            pred_labels += [label]
+            pred_values += [values]
+
+    if len(y) == 0:
+        y = [0] * nr_instance
+    ACC, MSE, SCC = evaluations(y, pred_labels)
+
+    if svm_type in [svm_forms.EPSILON_SVR, svm_forms.NU_SVR]:
+        info("Mean squared error = %g (regression)" % MSE)
+        info("Squared correlation coefficient = %g (regression)" % SCC)
+    else:
+        info("Accuracy = %g%% (%d/%d) (classification)" % (ACC, int(round(nr_instance*ACC/100)), nr_instance))
+
+    return pred_labels, (ACC, MSE, SCC), pred_values

libsvm-3.36/python/setup.py

@@ -0,0 +1,123 @@
+#!/usr/bin/env python
+
+import sys, os
+from os import path
+from shutil import copyfile, rmtree
+from glob import glob
+
+from setuptools import setup, Extension
+from distutils.command.clean import clean as clean_cmd
+
+# a technique to build a shared library on windows
+from distutils.command.build_ext import build_ext
+
+build_ext.get_export_symbols = lambda x, y: []
+
+
+PACKAGE_DIR = "libsvm"
+PACKAGE_NAME = "libsvm-official"
+VERSION = "3.36.0"
+cpp_dir = "cpp-source"
+# should be consistent with dynamic_lib_name in libsvm/svm.py
+dynamic_lib_name = "clib"
+
+# sources to be included to build the shared library
+source_codes = [
+    "svm.cpp",
+]
+headers = [
+    "svm.h",
+    "svm.def",
+]
+
+# license parameters
+license_source = path.join("..", "COPYRIGHT")
+license_file = "LICENSE"
+license_name = "BSD-3-Clause"
+
+kwargs_for_extension = {
+    "sources": [path.join(cpp_dir, f) for f in source_codes],
+    "depends": [path.join(cpp_dir, f) for f in headers],
+    "include_dirs": [cpp_dir],
+    "language": "c++",
+}
+
+# see ../Makefile.win and enable openmp
+if sys.platform == "win32":
+    kwargs_for_extension.update(
+        {
+            "define_macros": [("_WIN64", ""), ("_CRT_SECURE_NO_DEPRECATE", "")],
+            "extra_link_args": [r"-DEF:{}\svm.def".format(cpp_dir)],
+            "extra_compile_args": ["/openmp"],
+        }
+    )
+else:
+    kwargs_for_extension.update(
+        {
+            "extra_compile_args": ["-fopenmp"],
+            "extra_link_args": ["-fopenmp"],
+        }
+    )
+
+
+def create_cpp_source():
+    for f in source_codes + headers:
+        src_file = path.join("..", f)
+        tgt_file = path.join(cpp_dir, f)
+        # ensure blas directory is created
+        os.makedirs(path.dirname(tgt_file), exist_ok=True)
+        copyfile(src_file, tgt_file)
+
+
+class CleanCommand(clean_cmd):
+    def run(self):
+        clean_cmd.run(self)
+        to_be_removed = ["build/", "dist/", "MANIFEST", cpp_dir, "{}.egg-info".format(PACKAGE_NAME), license_file]
+        to_be_removed += glob("./{}/{}.*".format(PACKAGE_DIR, dynamic_lib_name))
+        for root, dirs, files in os.walk(os.curdir, topdown=False):
+            if "__pycache__" in dirs:
+                to_be_removed.append(path.join(root, "__pycache__"))
+            to_be_removed += [f for f in files if f.endswith(".pyc")]
+
+        for f in to_be_removed:
+            print("remove {}".format(f))
+            if f == ".":
+                continue
+            elif path.isfile(f):
+                os.remove(f)
+            elif path.isdir(f):
+                rmtree(f)
+
+def main():
+    if not path.exists(cpp_dir):
+        create_cpp_source()
+
+    if not path.exists(license_file):
+        copyfile(license_source, license_file)
+
+    with open("README") as f:
+        long_description = f.read()
+
+    setup(
+        name=PACKAGE_NAME,
+        packages=[PACKAGE_DIR],
+        version=VERSION,
+        description="Python binding of LIBSVM",
+        long_description=long_description,
+        long_description_content_type="text/plain",
+        author="ML group @ National Taiwan University",
+        author_email="cjlin@csie.ntu.edu.tw",
+        url="https://www.csie.ntu.edu.tw/~cjlin/libsvm",
+        license=license_name,
+        install_requires=["scipy"],
+        ext_modules=[
+            Extension(
+                "{}.{}".format(PACKAGE_DIR, dynamic_lib_name), **kwargs_for_extension
+            )
+        ],
+        cmdclass={"clean": CleanCommand},
+    )
+
+
+main()
+