#! /usr/bin/python

import numpy

import sys
import os
sys.path.append(os.path.join(os.path.abspath(os.path.dirname(__file__)),os.pardir))
import helperFunctions

import activeLearningGPLinKprototype

class Classifier(activeLearningGPLinKprototype.ClassifierPrototype):

  def __init__(self,
               sigmaN = 0.00178,
               useDensity = True,
               configFile = None):

    activeLearningGPLinKprototype.ClassifierPrototype.__init__(self, sigmaN=sigmaN, configFile=configFile)
    self.useDensity = helperFunctions.getConfig(configFile, 'activeLearning', 'useDensity', useDensity, 'bool', True)


  # x.shape = (number of samples, feat dim)
  def calcEMOC(self, x, allX=None, kAll=None, sigmaF=None):

    if allX is None:
      allX = numpy.append(self.X, x, axis=0)

    if kAll is None:
      kAll = numpy.dot(allX,allX.T)

    k = kAll[0:self.X.shape[0],self.X.shape[0]:]

    if sigmaF is None:
      selfK = numpy.asmatrix(numpy.diag(kAll[self.X.shape[0]:,self.X.shape[0]:])).T
      sigmaF = self.calcSigmaF(x, k, selfK)

    containsNoise = (self.yUni == -1).any()
    infY = self.infer(x, containsNoise, k)
    probs = self.calcProbs(x, infY, sigmaF)

    term1 = 1.0 / (self.sigmaN + sigmaF)

    term2 = numpy.asmatrix(numpy.ones((self.X.shape[0] + 1,x.shape[0]))) * (-1.0)
    term2[0:self.X.shape[0],:] = numpy.linalg.solve(numpy.add(self.K, numpy.identity(self.X.shape[0], dtype=numpy.float32)*self.sigmaN), k)

    pro = numpy.absolute(infY - 1)
    contra = numpy.absolute(infY + 1)

    term3 = numpy.repeat(numpy.sum(contra,axis=1),contra.shape[1],axis=1)
    term3 = numpy.add(numpy.subtract(term3,contra),pro)

    scores = numpy.asmatrix(numpy.zeros((x.shape[0],1)))
    for cls in range(infY.shape[1]):

      diffAlpha = numpy.multiply(numpy.repeat(numpy.multiply(term1,term3[:,cls]), term2.shape[0], axis=1),term2.T)

      change = numpy.dot(diffAlpha[:,:-1],kAll[0:self.X.shape[0],:])
      change = numpy.add(change, numpy.multiply(numpy.repeat(diffAlpha[:,-1], kAll.shape[0], axis=1),kAll[self.X.shape[0]:,:]))

      scores = numpy.add(scores, numpy.multiply(probs[:,cls],numpy.sum(numpy.absolute(change),axis=1)))

    return scores


  def getDensity(self, sim):
    return numpy.sum(sim, axis=1) / float(sim.shape[1])


  def getDiversity(self, sim):
    return 1.0 / numpy.max(sim, axis=1)


  # x.shape = (feat dim, number of samples)
  def calcAlScores(self, x):

    allX = numpy.append(self.X, x, axis=0)
    kAll = numpy.dot(allX,allX.T)

    k = kAll[0:self.X.shape[0],self.X.shape[0]:]
    selfK = numpy.asmatrix(numpy.diag(kAll[self.X.shape[0]:,self.X.shape[0]:])).T
    sigmaF = self.calcSigmaF(x, k, selfK)

    alScores = self.calcEMOC(x, allX=allX, kAll=kAll, sigmaF=sigmaF)

    if self.useDensity:

        density = self.getDensity(kAll[self.X.shape[0]:,:])
        alScores = numpy.multiply(alScores, density)

    return alScores