我目前正在使用来自触摸屏输入的数据作为 Behavioral Biometric来学习神经网络。基本上，我试图通过训练如下所示的神经网络模型来预测“用户 ID”。

import time
import os
BATCH_SIZE=32
embedding_size=256
sequence_length=200
BUFFER_SIZE=10000
input_size=41
learning_rate=0.001

inputs_as_tensors=tf.data.Dataset.from_tensor_slices(train_data_features_array)
targets_as_tensors=tf.data.Dataset.from_tensor_slices(train_data_labels_categorical_array)
training_data=tf.data.Dataset.zip((inputs_as_tensors,targets_as_tensors))
#training_data=training_data.batch(sequence_length,drop_remainder=True)
training_dataset=training_data.shuffle(BUFFER_SIZE).batch(BATCH_SIZE, drop_remainder=True)
print(training_dataset)

def build_model(vocab_size,batch_size):
    modelf=tf.keras.Sequential([
        tf.keras.layers.Dense(10,activation="sigmoid",input_shape=(None,10)),
        
                                tf.keras.layers.Dense(30,activation="relu",use_bias=True),  
                                 
                                     tf.keras.layers.Dropout(0.2),                    
                                   tf.keras.layers.Dense(vocab_size)
                          
                               
                                ])
    return modelf

def training_step(inputs,targets,optimizer):
    with tf.GradientTape() as tape:
        predictions=model(inputs)
        loss=tf.reduce_mean(tf.keras.losses.categorical_crossentropy(targets,predictions,from_logits=True))
        
        grads=tape.gradient(loss,model.trainable_variables)
        optimizer.apply_gradients(zip(grads,model.trainable_variables))
        return loss,predictions

model=build_model(input_size,BATCH_SIZE)
i=0
inner_loop=0
checkpoint_dir ='Moses_Model_x'
checkpoint_prefix = os.path.join(checkpoint_dir, "ckpt_{i}")

while(1):
    start = time.time()
    for x,y in training_dataset:
        loss,predictions=training_step(x,y,tf.keras.optimizers.RMSprop(learning_rate=0.002))
    print ('Epoch {} Loss {:.4f}'.format(i, loss))
    print ('Time taken for iteration {} is {} sec\n'.format(i,time.time() - start))
    model.save_weights(checkpoint_prefix.format(i=i))
    i=i+1

但是，损失值实际上是在增加的。我的代码中有什么需要更改或有什么问题吗？