Created by: 357589873

数据：一些网页的title提取后取汉语部分。每个分类10W+的量 class:三分类 网络结构：

def stacked_lstm_net(input_dim,
                     class_dim=2,
                     emb_dim=128,
                     hid_dim=512,
                     stacked_num=5):    

    assert stacked_num % 2 == 1
    fc_para_attr = paddle.attr.Param(learning_rate=1e-3)
    lstm_para_attr = paddle.attr.Param(initial_std=0., learning_rate=1.)
    para_attr = [fc_para_attr, lstm_para_attr]
    bias_attr = paddle.attr.Param(initial_std=0., l2_rate=0.)
    relu = paddle.activation.Relu()
    brelu = paddle.activation.BRelu()
    tanh = paddle.activation.Tanh()
    linear = paddle.activation.Linear()

    data = paddle.layer.data("x",
                             paddle.data_type.integer_value_sequence(input_dim))
    emb = paddle.layer.embedding(input=data, size=emb_dim)
    fc1 = paddle.layer.fc(
        input=emb, size=hid_dim, act=brelu, bias_attr=bias_attr)
    lstm1 = paddle.layer.lstmemory(input=fc1, act=tanh, bias_attr=bias_attr)

    inputs = [fc1, lstm1]
    for i in range(2, stacked_num + 1):
        fc = paddle.layer.fc(
            input=inputs,
            size=hid_dim,
            act=brelu,
            param_attr=para_attr,
            bias_attr=bias_attr)
        lstm = paddle.layer.lstmemory(
            input=fc, reverse=(i % 2) == 0, act=tanh, bias_attr=bias_attr)
        inputs = [fc, lstm]

    fc_last = paddle.layer.pooling(
        input=inputs[0], pooling_type=paddle.pooling.Max())
    lstm_last = paddle.layer.pooling(
        input=inputs[1], pooling_type=paddle.pooling.Max())
    output = paddle.layer.fc(
        input=[fc_last, lstm_last],
        size=class_dim,
        act=paddle.activation.Softmax(),

        layer_attr=paddle.attr.ExtraLayerAttribute(
            error_clipping_threshold=10.0
        ),
        bias_attr=bias_attr,
        param_attr=para_attr)

    lbl = paddle.layer.data("y", paddle.data_type.integer_value(1))
    cost = paddle.layer.classification_cost(input=output, label=lbl)
    return cost, output, lbl

参数选择：

cost, prob, label = stacked_lstm_net(
        input_dim=len(meta[0]),
        class_dim=3,
        emb_dim=word_dim,
        hid_dim=512,
        stacked_num=3,
    )
 optimizer = paddle.optimizer.Adam(
        learning_rate=0.00001,
        gradient_clipping_threshold=2,
        regularization=paddle.optimizer.L2Regularization(rate=0.00001),
        model_average=paddle.optimizer.ModelAverage(average_window=0.5)
    )
 word_dim = min(meta[-1], 200)

调整了learning_rate 和gradient_clipping_threshold 没起作用。