我在训练一个神经网络，并得到了以下输出。损失和验证损失都在下降，这让我很高兴。然而，验证准确率保持不变。这可能是什么原因呢？我的数据相当不平衡，但我使用sklearn的compute_class_weight函数对其进行了加权处理。

Train on 109056 samples, validate on 27136 samplesEpoch 1/200- 1174s - loss: 1.0353 - acc: 0.5843 - val_loss: 1.0749 - val_acc: 0.7871Epoch 00001: val_acc improved from -inf to 0.78711, saving model to nn_best_weights.h5Epoch 2/200- 1174s - loss: 1.0122 - acc: 0.6001 - val_loss: 1.0642 - val_acc: 0.9084Epoch 00002: val_acc improved from 0.78711 to 0.90842, saving model to nn_best_weights.h5Epoch 3/200- 1176s - loss: 0.9974 - acc: 0.5885 - val_loss: 1.0445 - val_acc: 0.9257Epoch 00003: val_acc improved from 0.90842 to 0.92571, saving model to nn_best_weights.h5Epoch 4/200- 1177s - loss: 0.9834 - acc: 0.5760 - val_loss: 1.0071 - val_acc: 0.9260Epoch 00004: val_acc improved from 0.92571 to 0.92597, saving model to nn_best_weights.h5Epoch 5/200- 1182s - loss: 0.9688 - acc: 0.5639 - val_loss: 1.0175 - val_acc: 0.9260Epoch 00005: val_acc did not improve from 0.92597Epoch 6/200- 1177s - loss: 0.9449 - acc: 0.5602 - val_loss: 0.9976 - val_acc: 0.9246Epoch 00006: val_acc did not improve from 0.92597Epoch 7/200- 1186s - loss: 0.9070 - acc: 0.5598 - val_loss: 0.9667 - val_acc: 0.9258Epoch 00007: val_acc did not improve from 0.92597Epoch 8/200- 1178s - loss: 0.8541 - acc: 0.5663 - val_loss: 0.9254 - val_acc: 0.9221Epoch 00008: val_acc did not improve from 0.92597Epoch 9/200- 1171s - loss: 0.7859 - acc: 0.5853 - val_loss: 0.8686 - val_acc: 0.9237Epoch 00009: val_acc did not improve from 0.92597Epoch 10/200- 1172s - loss: 0.7161 - acc: 0.6139 - val_loss: 0.8119 - val_acc: 0.9260Epoch 00010: val_acc did not improve from 0.92597Epoch 11/200- 1168s - loss: 0.6500 - acc: 0.6416 - val_loss: 0.7531 - val_acc: 0.9259Epoch 00011: val_acc did not improve from 0.92597Epoch 12/200- 1164s - loss: 0.5967 - acc: 0.6676 - val_loss: 0.7904 - val_acc: 0.9260Epoch 00012: val_acc did not improve from 0.92597Epoch 13/200- 1175s - loss: 0.5608 - acc: 0.6848 - val_loss: 0.7589 - val_acc: 0.9259Epoch 00013: val_acc did not improve from 0.92597Epoch 14/200- 1221s - loss: 0.5377 - acc: 0.6980 - val_loss: 0.7811 - val_acc: 0.9260Epoch 00014: val_acc did not improve from 0.92597

我的模型如下所示。我知道卷积核大小相当大，但这是有意为之，因为数据以某种方式结构化。

    cnn = Sequential()    cnn.add(Conv2D(16, kernel_size=(2, 100), padding='same', data_format="channels_first", input_shape=(1,10, 100)))    cnn.add(LeakyReLU(alpha=0.01))    cnn.add(BatchNormalization())    cnn.add(Conv2D(16, (1, 1)))    cnn.add(LeakyReLU(alpha=0.01))    cnn.add(Conv2D(16, (1, 8)))    cnn.add(Flatten())     rnn = Sequential()    rnn = LSTM(100, return_sequences=False, dropout=0.2)    dense = Sequential()    dense.add(Dense(3, activation='softmax'))    main_input = Input(batch_shape=(512, 1, 1, 10, 100))     model = TimeDistributed(cnn)(main_input)     model = rnn(model)     model = dense(model)     replica = Model(inputs=main_input, outputs=model)    replica.compile(loss='categorical_crossentropy', optimizer='adam',  metrics=['accuracy'])

回答：

不了解你的模型，很难回答你的问题。

可能的答案包括：

• 你的模型没有问题。这可能是你能达到的最高准确率。
• 你的数据可能不平衡或未经洗牌。验证准确率高于训练准确率可能表明训练、评估和测试数据的分割存在问题。训练准确率通常在开始时高于验证准确率。然后验证准确率会赶上，或者不会。这也可能表明你的数据集变化不大，因此可能出现这种情况。
• 你的学习率可能过大。尝试将其降低。

我猜测模型实际优化的指标是损失，因此在优化过程中你应该关注损失并监控其改进情况。

查看此链接以获取更多关于如何检查模型的信息。