loss不下降

[luoshizhi]

您好，小白请教，我用你的ensemble.py跑花呗的数据，loss下降到0.47左右就下不去了，请问可能是哪的问题呢？

Total params: 1,580,136
Trainable params: 933,889
Non-trainable params: 646,247
__________________________________________________________________________________________________
Epoch 1/15
147653/147653 [==============================] - 24s 165us/step - loss: 0.5730
Epoch 2/15
147653/147653 [==============================] - 23s 153us/step - loss: 0.4814
Epoch 3/15
147653/147653 [==============================] - 23s 154us/step - loss: 0.4700
Epoch 4/15
147653/147653 [==============================] - 23s 154us/step - loss: 0.4695
Epoch 5/15
147653/147653 [==============================] - 23s 154us/step - loss: 0.4695
Epoch 6/15
147653/147653 [==============================] - 23s 154us/step - loss: 0.4695
Epoch 7/15
147653/147653 [==============================] - 23s 153us/step - loss: 0.4695
Epoch 8/15
147653/147653 [==============================] - 23s 153us/step - loss: 0.4695
Epoch 9/15
147653/147653 [==============================] - 23s 154us/step - loss: 0.4695
Epoch 10/15
147653/147653 [==============================] - 23s 153us/step - loss: 0.4695
Epoch 11/15
147653/147653 [==============================] - 23s 154us/step - loss: 0.4695
Epoch 12/15
147653/147653 [==============================] - 23s 154us/step - loss: 0.4694
Epoch 13/15
147653/147653 [==============================] - 23s 154us/step - loss: 0.4695
Epoch 14/15
147653/147653 [==============================] - 23s 154us/step - loss: 0.4695
Epoch 15/15
147653/147653 [==============================] - 23s 155us/step - loss: 0.4695
iter 1

[raven4752]

您好，请问您使用的是什么数据集和词向量？做了哪些修改呢？