def eval_clf_accuracy()

in src/evaluation.py [0:0]

• 23 lines of code
• 8 McCabe index (conditional complexity)

    def eval_clf_accuracy(self):
        """
        Compute the accuracy of flipped attributes according to the trained classifier.
        """
        data = self.data
        params = self.params
        self.ae.eval()
        bs = params.batch_size

        idx = []
        for j in range(len(params.attr)):
            attr_index = self.eval_clf.attr.index(params.attr[j])
            idx.append(sum([x[1] for x in self.eval_clf.attr[:attr_index]]))

        all_preds = [[] for _ in range(params.n_attr)]
        for i in range(0, len(data), bs):
            # batch / encode / decode
            batch_x, batch_y = data.eval_batch(i, i + bs)
            enc_outputs = self.ae.encode(batch_x)
            # flip all attributes one by one
            k = 0
            for j, (_, n_cat) in enumerate(params.attr):
                for value in range(n_cat):
                    flipped = flip_attributes(batch_y, params, j, new_value=value)
                    dec_outputs = self.ae.decode(enc_outputs, flipped)
                    # classify
                    clf_preds = self.eval_clf(dec_outputs[-1])[:, idx[j]:idx[j] + n_cat].max(1)[1].view(-1)
                    all_preds[k].extend((clf_preds.data.cpu() == value).tolist())
                    k += 1
            assert k == params.n_attr

        return [np.mean(x) for x in all_preds]