Question

我经常打成这样，

ValueError: Input 0 of layer sequential is incompatible with the layer: expected ndim=3, found ndim=2. Full shape received: [10 ,3]

我四处搜寻，发现了

LSTM layer expects inputs to have shape of (batch_size, timesteps, input_dim)

好的，但是老实说我还是有些困惑。

例如，我有这样的训练数据

x_train (100,3) #it consists of like `[[1,2,3],[3,4,5],[5,6,7]]`
y_train (100,3) #answers

我想使用10组3对数字并预测[7,8,9]下的3对数字。

从x_train [1〜10]到y_train [11]一样的猜测

下面的代码有效，但是，我仍然不清楚

在input_shape=(3,1)处1的含义是什么？？。应该是3（我要最终确定的尺寸）

batch_size是LSTM请求的第一个参数。

所以，当我想从过去10个项目中预测一项时，在此处设置10是否正确？

x_train = np.array(x).reshape(100, 3,1)
y_train = np.array(x).reshape(100, 3,1)

model.add(LSTM(512, activation=None, input_shape=(3, 1), return_sequences=True))

model.add(Dense(1, activation="linear"))

opt = Adam(lr=0.001)

model.compile(loss='mse', optimizer=opt)
model.summary()
history = model.fit(x_train, y_train, epochs=epoch, batch_size=10) // how to set batch size???

Answer 1

尝试以下代码：

import tensorflow as tf
import numpy as np
x = np.random.uniform(0, 10, [101, 3])

x_train = np.array(x[:-1]).reshape(-1, 5, 3) # your data comprise of 20 sequences
y_train = np.array(x[1:]).reshape(-1, 5, 3)

model = tf.keras.Sequential()
model.add(tf.keras.layers.LSTM(512, activation=None, input_shape=(None, 3), return_sequences=True))

model.add(tf.keras.layers.Dense(1, activation="linear"))

opt = tf.keras.optimizers.Adam(lr=0.001)

model.compile(loss='mse', optimizer=opt)
model.summary()
history = model.fit(x_train, y_train, epochs=10, batch_size=10) # here you can set a batch size (your 20 sequences will be splitted into two batches)

如何设置keras LSTM的输入形状

1 个答案: