这是我遇到的错误 –

TypeError                                 Traceback (most recent call last)<ipython-input-22-512cdec6aaaf> in <module>()      7     y_pred=model_eval.fit(x_train,y_train,validation_data=(x_test,y_test),epochs=epochs)      8     y_pred_labels=np.argmax(y_pred, axis=-1)----> 9     confusion_matrix = metrics.confusion_matrix(y_true=y_train_labels, y_pred=y_pred_labels)  # shape=(12, 12)4 frames/usr/local/lib/python3.6/dist-packages/sklearn/utils/validation.py in _num_samples(x)    144         if len(x.shape) == 0:    145             raise TypeError("Singleton array %r cannot be considered"--> 146                             " a valid collection." % x)    147         # Check that shape is returning an integer or default to len    148         # Dask dataframes may not return numeric shape[0] valueTypeError: Singleton array 0 cannot be considered a valid collection.

这也可能是一种低效或错误的混淆矩阵编码方式。如果是这样，请指导我走上正确的道路。

回答：

将y_train_labels和y_pred_labels都重塑为(N,1)而不是(N,)：

y_train_labels = y_train_labels.reshape(-1,1)y_pred_labels  = y_pred_labels.reshape(-1,1)