mlp.py 文件源码

def evaluateClassifier(self, dataset, batch_size):
        """
                         Evaluate neg log likelihood and accuracy
        """
        N = dataset['X'].shape[0]
        crossentropy = 0
        ncorrect = 0
        for bnum, st_idx in enumerate(range(0,N,batch_size)):
            end_idx = min(st_idx+batch_size,N)
            X = self.sampleDataset(dataset['X'][st_idx:end_idx].astype(config.floatX))
            Y = dataset['Y'][st_idx:end_idx].astype('int32')
            batch_crossentropy, batch_ncorrect = self.evaluate(X=X,Y=Y)
            crossentropy += batch_crossentropy
            ncorrect += batch_ncorrect
        crossentropy /= float(N)
        accuracy = ncorrect/float(N)
        return crossentropy, accuracy