Question

我尝试在张量流上进行多标签分类，但性能不佳。

输入：[1,0,1，...，0]（单热）和文字内容

   with tf.name_scope("rnn"):
        cells = [dropout() for _ in range(self.config.num_layers)]
        rnn_cell = tf.contrib.rnn.MultiRNNCell(cells, state_is_tuple=True)
        _outputs, _ = tf.nn.dynamic_rnn(cell=rnn_cell, inputs=embedding_inputs, dtype=tf.float32)
        last = _outputs[:, -1, :]

模型：嵌入层

   with tf.name_scope("score"):

        fc = tf.layers.dense(last, self.config.hidden_dim, name='fc1')
        fc = tf.contrib.layers.dropout(fc, self.keep_prob)
        fc = tf.nn.relu(fc)

        self.logits = tf.layers.dense(fc, self.config.num_classes, name='fc2')
        self.y_pred_cls = tf.nn.sigmoid(self.logits, name='pred')

双LSTM

   with tf.name_scope("optimize"):

        cross_entropy = tf.nn.sigmoid_cross_entropy_with_logits(logits=self.logits, labels=self.input_y)
        self.loss = tf.reduce_mean(cross_entropy)
        self.optim = tf.train.AdamOptimizer(learning_rate=self.config.learning_rate).minimize(self.loss)

通过relu

激活

    with tf.name_scope("accuracy"):

        threashold=0.5
        self.labels = tf.equal(self.input_y, 1)
        self.preds = tf.greater_equal(self.y_pred_cls,threashold)
        self.correct_num = tf.divide(tf.size(tf.where(tf.logical_not(tf.logical_xor(self.labels, self.preds)))),2)
        self.total_num = tf.size(self.input_y)
        self.correct_pred = tf.divide(tf.cast(self.correct_num, tf.float32),tf.cast(self.total_num, tf.float32))

        self.acc = tf.reduce_mean(tf.cast(self.correct_pred, tf.float32))

使用sigmoid cross entropy

$a = 'I need the 11 Stagione';
if ( preg_match( "/\b(Season|Stagione|S|T)\s*\d+|\d+\s*(?1)\b/i", $a, $matches ) )
{
    print_r( $matches[0] );
}

计算acc

\b

模型预测[0,0，.... 0]（全为零）对测试数据，不属于任何类别。

我的代码或模型设计出了什么问题？谢谢。

张量流中性能差的多标签分类

0 个答案: