04b4eb0d841f8dfe1d095b2af4bcf57cda72c1d2,lxmls/classifiers/multinomial_naive_bayes.py,MultinomialNaiveBayes,train,#MultinomialNaiveBayes#Any#Any#,16

Before Change

            likelihood[:,i] = value // likelihood = count of occurrences of a word in a class
                                    // NOTE: at this point this is a count, not a likelihood
                
        for f in xrange(nr_f):
            for i in xrange(nr_c):
                if self.smooth:
                    likelihood[f,i] = (self.smooth_param + likelihood[f,i])/(nr_f*self.smooth_param + sums[f,0]) // Add-one smoothing
                else:
                    likelihood[f,i] = likelihood[f,i]/sums[f,0] 

        params = np.zeros((nr_f+1,nr_c))
        for i in xrange(nr_c):
            params[0,i] = np.log(prior[i])
            params[1:,i] = np.nan_to_num(np.log(likelihood[:,i]))

After Change

            prior[i] = 1.0*len(idx)/len(y)
            // compute word counts for this class
            class_count[:,i] = sum(x[idx,:],0)
            total_count[:,0] += sum(x[idx,:],0)

	// Compute likelihood from counts, special
        if self.smooth:
            likelihood = (self.smooth_param + class_count)/(total_count + nr_f*self.smooth_param) 	

Instances