error

[9650X]

The token $end should be one of those: ASSERT, AT, BACKQUOTE, BINARY, BINARY_RAW_STRING, BINARY_STRING, BREAK, CLASS, COMMENT, COMPLEX, CONTINUE, DEDENT, DEF, DEL, ELLIPSIS, ENDL, ENDMARKER, FLOAT, FLOAT_EXPONANT, FLOAT_EXPONANT_COMPLEX, FOR, FROM, GLOBAL, HEXA, IF, IMPORT, INT, INTERPOLATED_RAW_STRING, INTERPOLATED_STRING, LAMBDA, LEFT_BRACKET, LEFT_PARENTHESIS, LEFT_SQUARE_BRACKET, LONG, MINUS, NAME, NONLOCAL, NOT, OCTA, PASS, PLUS, RAISE, RAW_STRING, RETURN, STAR, STRING, TILDE, TRY, UNICODE_RAW_STRING, UNICODE_STRING, WHILE, WITH, YIELD

Baron has failed to parse this input. If this is valid python code (and by that I mean that the python binary successfully parse this code without any syntax error) (also consider that baron does not yet parse python 3 code integrally) it would be kind if you can extract a snippet of your code that make Baron fails and open a bug here: https://github.com/PyCQA/baron/issues

Sorry for the inconvenience.

Tasks

Preview Give feedback

No tasks being tracked yet.

[9650X]

from abc import abstractmethod
from numpy import eye, concatenate, zeros, shape, mean, reshape, arange, exp, outer,
linalg, dot, cos, sin, sqrt, inf
from numpy.random import permutation
from numpy.lib.index_tricks import fill_diagonal
from matplotlib.pyplot import imshow,show
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.cm as cm
import warnings

from kerpy_master.tools.GenericTests import GenericTests

class Kernel(object):
def init(self):
self.rff_num=None
self.rff_freq=None
pass

def __str__(self):
    s=""
    return s

@abstractmethod
def kernel(self, X, Y=None):
    raise NotImplementedError()

@abstractmethod
def set_kerpar(self,kerpar):
    self.set_width(kerpar)

@abstractmethod
def set_width(self, width):
    if hasattr(self, 'width'):
        warnmsg="\nChanging kernel width from "+str(self.width)+" to "+str(width)
        #warnings.warn(warnmsg) ---need to add verbose argument to show these warning messages
        if self.rff_freq is not None:
            warnmsg="\nrff frequencies found. rescaling to width " +str(width)
            #warnings.warn(warnmsg)
            self.rff_freq=self.unit_rff_freq/width
        self.width=width
    else:
        raise ValueError("Senseless: kernel has no 'width' attribute!")

@abstractmethod
def rff_generate(self,m,dim=1):
    raise NotImplementedError()

@abstractmethod
def rff_expand(self,X):
    if self.rff_freq is None:
        raise ValueError("rff_freq has not been set. use rff_generate first")
    """
    Computes the random Fourier features for the input dataset X
    for a set of frequencies in rff_freq.
    This set of frequencies has to be precomputed
    X - 2d numpy.ndarray, first set of samples:
        number of rows: number of samples
        number of columns: dimensionality
    """
    GenericTests.check_type(X, 'X',np.ndarray)
    xdotw=dot(X,(self.rff_freq).T)
    return sqrt(2./self.rff_num)*np.concatenate( ( cos(xdotw),sin(xdotw) ) , axis=1 )
    
@abstractmethod
def gradient(self, x, Y):
    
    # ensure this in every implementation
    assert(len(shape(x))==1)
    assert(len(shape(Y))==2)
    assert(len(x)==shape(Y)[1])
    
    raise NotImplementedError()

@staticmethod
def centering_matrix(n):
    """
    Returns the centering matrix eye(n) - 1.0 / n
    """
    return eye(n) - 1.0 / n

@staticmethod
def center_kernel_matrix(K):
    """
    Centers the kernel matrix via a centering matrix H=I-1/n and returns HKH
    """
    n = shape(K)[0]
    H = eye(n) - 1.0 / n
    return  1.0 / n * H.dot(K.dot(H))


@abstractmethod
def show_kernel_matrix(self,X,Y=None):
    K=self.kernel(X,Y)
    imshow(K, interpolation="nearest")
    show()

@abstractmethod
def svc(self,X,y,lmbda=1.0,Xtst=None,ytst=None):
    from sklearn import svm
    svc=svm.SVC(kernel=self.kernel,C=lmbda)
    svc.fit(X,y)
    if Xtst is None:
        return svc
    else:
        ypre=svc.predict(Xtst)
        if ytst is None:
            return svc,ypre
        else:
            return svc,ypre,1-svc.score(Xtst,ytst)

@abstractmethod
def svc_rff(self,X,y,lmbda=1.0,Xtst=None,ytst=None):
    from sklearn import svm
    phi=self.rff_expand(X)
    svc=svm.LinearSVC(C=lmbda,dual=True)
    svc.fit(phi,y)
    if Xtst is None:
        return svc
    else:
        phitst=self.rff_expand(Xtst)
        ypre=svc.predict(phitst)
        if ytst is None:
            return svc,ypre
        else:
            return svc,ypre,1-svc.score(phitst,ytst)

@abstractmethod
def ridge_regress(self,X,y,lmbda=0.01,Xtst=None,ytst=None):
    K=self.kernel(X)
    n=shape(K)[0]
    aa=linalg.solve(K+lmbda*eye(n),y)
    if Xtst is None:
        return aa
    else:
        ypre=dot(aa.T,self.kernel(X,Xtst)).T
        if ytst is None:
            return aa,ypre
        else:
            return aa,ypre,(linalg.norm(ytst-ypre)**2)/np.shape(ytst)[0]

@abstractmethod
def ridge_regress_rff(self,X,y,lmbda=0.01,Xtst=None,ytst=None):

if self.rff_freq is None:

warnings.warn("\nrff_freq has not been set!\nGenerating new random frequencies (m=100 by default)")

self.rff_generate(100,dim=shape(X)[1])

print shape(X)[1]

    phi=self.rff_expand(X)
    bb=linalg.solve(dot(phi.T,phi)+lmbda*eye(self.rff_num),dot(phi.T,y))
    if Xtst is None:
        return bb
    else:
        phitst=self.rff_expand(Xtst)
        ypre=dot(phitst,bb)
        if ytst is None:
            return bb,ypre
        else:
            return bb,ypre,(linalg.norm(ytst-ypre)**2)/np.shape(ytst)[0]

@abstractmethod
def xvalidate( self,X,y, method = 'ridge_regress',regpar_grid=(1+arange(25))/200.0, kerpar_grid=exp(-13+arange(25)), numFolds = 10, verbose = False, visualise = False):
    from sklearn import model_selection
    which_method = getattr(self,method)
    n=len(X)
    kf=model_selection.KFold(n,n_folds=numFolds)
    xvalerr=zeros((len(regpar_grid),len(kerpar_grid)))
    width_idx=0
    for width in kerpar_grid:
        try:
            self.set_kerpar(width)
        except ValueError:
            xvalerr[:,width_idx]=inf
            warnings.warn("...invalid kernel parameter value in cross-validation. ignoring\n")
            width_idx+=1
            continue
        else:
            lmbda_idx=0
            for lmbda in regpar_grid:
                fold = 0
                prederr = zeros(numFolds)
                for train_index, test_index in kf:
                    if type(X)==list:
                        #could use slicing to speed this up when X is a list
                        #currently uses sklearn cross_validation framework which returns indices as arrays
                        #so simple list comprehension below
                        X_train = [X[i] for i in train_index]
                        X_test = [X[i] for i in test_index]
                    else:
                        X_train, X_test = X[train_index], X[test_index]
                    if type(y)==list:
                        y_train = [y[i] for i in train_index]
                        y_test = [y[i] for i in test_index]
                    else:
                        y_train, y_test = y[train_index], y[test_index]
                    _,_,prederr[fold]=which_method(X_train,y_train,lmbda=lmbda,Xtst=X_test,ytst=y_test)
                    fold+=1
                xvalerr[lmbda_idx,width_idx]=mean(prederr)
                if verbose:
                    print("kerpar:"+str(width)+", regpar:"+str(lmbda))
                    print("    cross-validated loss:"+str(xvalerr[lmbda_idx,width_idx]))
                lmbda_idx+=1
            width_idx+=1
    min_idx = np.unravel_index(np.argmin(xvalerr),shape(xvalerr))
    if visualise:
        plt.imshow(xvalerr, interpolation='none',
            origin='lower',
            cmap=cm.pink)
            #extent=(regpar_grid[0],regpar_grid[-1],kerpar_grid[0],kerpar_grid[-1]))
        plt.colorbar()
        plt.title("cross-validated loss")
        plt.ylabel("regularisation parameter")
        plt.xlabel("kernel parameter")
        show()
    return regpar_grid[min_idx[0]],kerpar_grid[min_idx[1]]

@abstractmethod
def estimateMMD(self,sample1,sample2,unbiased=False):
    """
    Compute the MMD between two samples
    """
    K11 = self.kernel(sample1)
    K22 = self.kernel(sample2)
    K12 = self.kernel(sample1,sample2)
    if unbiased:
        fill_diagonal(K11,0.0)
        fill_diagonal(K22,0.0)
        n=float(shape(K11)[0])
        m=float(shape(K22)[0])
        return sum(sum(K11))/(pow(n,2)-n) + sum(sum(K22))/(pow(m,2)-m) - 2*mean(K12[:])
    else:
        return mean(K11[:])+mean(K22[:])-2*mean(K12[:])
        
@abstractmethod
def estimateMMD_rff(self, sample1, sample2, unbiased=False):
  phi1 = self.rff_expand(sample1)
  phi2 = self.rff_expand(sample2)
  featuremean1 = mean(phi1, axis=0)
  featuremean2 = mean(phi2, axis=0)
  if unbiased:
      nx = shape(phi1)[0]
      ny = shape(phi2)[0]
      first_term = nx / (nx - 1.0) * (dot(featuremean1, featuremean1) - mean(linalg.norm(phi1, axis=1)**2) / nx)
      second_term = ny / (ny - 1.0) * (dot(featuremean2, featuremean2) - mean(linalg.norm(phi2, axis=1)**2) / ny)
      third_term = -2 * dot(featuremean1, featuremean2)
      return first_term + second_term + third_term
  else:
      return linalg.norm(featuremean1 - featuremean2)**2

代码报错的地方在最后一行。