代码

Question

我正在尝试在Rstudio Keras中构建用于文本分类的CNN。我是神经网络世界的新手，所以我只是看一些例子，看看我如何从别人的行为中学到东西。我在下面的代码中使用的模型规范是从我发现on the package GitHub.的示例中复制的。我将进一步参考此代码/示例作为“示例”。

当我按“原样”运行示例时（即，使用包作者提供的数据），它运行正常。但是，当我使用自己的数据（提供here in rdata format）时，模型不会编译。它卡在以下错误消息上：

py_call_impl中的错误（callable，dots $ args，dots $ keywords）： ValueError：检查目标时出错：期望dense_22有形状（无，10）但是有形状的数组（28686,11）

对我而言，这意味着y_train（即类别标签）的形状错误。但是，当我将标签的尺寸和数据类型与示例的尺寸和数据类型进行比较时，它们具有相似的格式。两者都是矩阵，其中行代表训练案例，而cols是表示标签的单热矢量。我的嵌入矩阵（embedding_matrix）和我的单词序列（x_test, x_train）

也是如此

有没有人对这里可能出现的问题有任何想法？

代码

library(keras)
library(stringi)
MAX_SEQUENCE_LENGTH <- 1000
MAX_NB_WORDS <- 20000
EMBEDDING_DIM <- 100

#Load my data
load("testdata.rdata")

#MARK start code copied from example
#without any changes

# load pre-trained word embeddings into an Embedding layer
# note that we set trainable = False so as to keep the embeddings fixed
num_words <- min(MAX_NB_WORDS, length(word_index))
 embedding_layer <- layer_embedding(
  input_dim = num_words,
  output_dim = EMBEDDING_DIM,
  weights = list(embedding_matrix),
  input_length = MAX_SEQUENCE_LENGTH,
  trainable = FALSE
)

cat('Training model\n')

# train a 1D convnet with global maxpooling
sequence_input <- layer_input(shape = list(MAX_SEQUENCE_LENGTH), dtype='int32')

preds <- sequence_input %>%
  embedding_layer %>% 
  layer_conv_1d(filters = 128, kernel_size = 5, activation = 'relu') %>% 
  layer_max_pooling_1d(pool_size = 5) %>% 
  layer_conv_1d(filters = 128, kernel_size = 5, activation = 'relu') %>% 
  layer_max_pooling_1d(pool_size = 5) %>% 
  layer_conv_1d(filters = 128, kernel_size = 5, activation = 'relu') %>% 
  layer_max_pooling_1d(pool_size = 35) %>% 
  layer_flatten() %>% 
  layer_dense(units = 128, activation = 'relu') %>% 
  layer_dense(units = length(labels_index), activation = 'softmax')


model <- keras_model(sequence_input, preds)

model %>% compile(
  loss = 'categorical_crossentropy',
  optimizer = 'rmsprop',
  metrics = c('acc')  
)

model %>% fit(
  x_train, y_train,
  batch_size = 128,
  epochs = 10,
  validation_data = list(x_val, y_val)
)

模型摘要：

________________________________________________________________________________
Layer (type)                        Output Shape                    Param #     
================================================================================
input_2 (InputLayer)                (None, 1000)                    0           
________________________________________________________________________________
embedding_2 (Embedding)             (None, 1000, 100)               2000000     
________________________________________________________________________________
conv1d_4 (Conv1D)                   (None, 996, 128)                64128       
________________________________________________________________________________
max_pooling1d_4 (MaxPooling1D)      (None, 199, 128)                0           
________________________________________________________________________________
conv1d_5 (Conv1D)                   (None, 195, 128)                82048       
________________________________________________________________________________
max_pooling1d_5 (MaxPooling1D)      (None, 39, 128)                 0           
________________________________________________________________________________
conv1d_6 (Conv1D)                   (None, 35, 128)                 82048       
________________________________________________________________________________
max_pooling1d_6 (MaxPooling1D)      (None, 1, 128)                  0           
________________________________________________________________________________
flatten_2 (Flatten)                 (None, 128)                     0           
________________________________________________________________________________
dense_3 (Dense)                     (None, 128)                     16512       
________________________________________________________________________________
dense_4 (Dense)                     (None, 20)                      2580        
================================================================================
Total params: 2,247,316
Trainable params: 247,316
Non-trainable params: 2,000,000
________________________________________________________________________________

Answer 1

不确定Glove数据维度，但您自己的y_train确实有11列而不是10列：

load("testdata.rdata")
dim(y_train)
# [1] 28686    11

此外，您的第一列V1似乎是虚假的，因为它只包含零：

> summary(y_train) # shown only V1-V8
       V1          V2               V3                V4               V5                V6               V7                V8         
 Min.   :0   Min.   :0.0000   Min.   :0.00000   Min.   :0.0000   Min.   :0.00000   Min.   :0.0000   Min.   :0.00000   Min.   :0.00000  
 1st Qu.:0   1st Qu.:0.0000   1st Qu.:0.00000   1st Qu.:0.0000   1st Qu.:0.00000   1st Qu.:0.0000   1st Qu.:0.00000   1st Qu.:0.00000  
 Median :0   Median :0.0000   Median :0.00000   Median :0.0000   Median :0.00000   Median :0.0000   Median :0.00000   Median :0.00000  
 Mean   :0   Mean   :0.4336   Mean   :0.01928   Mean   :0.1167   Mean   :0.08143   Mean   :0.1171   Mean   :0.06041   Mean   :0.07227  
 3rd Qu.:0   3rd Qu.:1.0000   3rd Qu.:0.00000   3rd Qu.:0.0000   3rd Qu.:0.00000   3rd Qu.:0.0000   3rd Qu.:0.00000   3rd Qu.:0.00000  
 Max.   :0   Max.   :1.0000   Max.   :1.00000   Max.   :1.0000   Max.   :1.00000   Max.   :1.0000   Max.   :1.00000   Max.   :1.00000

因此，如果10确实是目标维度（您的labels_index确实长度为10），则从y_train中删除（无用的）第一列应解决错误：

y_train_new <- y_train[,2:11]
dim(y_train_new)
# [1] 28686    10

R keras“ValueError：检查目标时出错”，代码与其他数据一起使用

代码

模型摘要：

1 个答案: