Блог

Я обучаю свою модель, чтобы классифицировать этапы ожидания, после извлечения объектов из моего сигнала я собрал объекты (X) в фрейме данных с формой (335,48) и y (метки) в форме (335,)

это мой код :

 def get_base_model():
    inp = Input(shape=(335,48))
    img_1 = Convolution1D(16, kernel_size=5, activation=activations.relu, padding="valid")(inp)
    img_1 = Convolution1D(16, kernel_size=5, activation=activations.relu, padding="valid")(img_1)
    img_1 = MaxPool1D(pool_size=2)(img_1)
    img_1 = SpatialDropout1D(rate=0.01)(img_1)
    img_1 = Convolution1D(32, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = Convolution1D(32, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = MaxPool1D(pool_size=2)(img_1)
    img_1 = SpatialDropout1D(rate=0.01)(img_1)
    img_1 = Convolution1D(32, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = Convolution1D(32, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = MaxPool1D(pool_size=2)(img_1)
    img_1 = SpatialDropout1D(rate=0.01)(img_1)
    img_1 = Convolution1D(256, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = Convolution1D(256, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = GlobalMaxPool1D()(img_1)
    img_1 = Dropout(rate=0.01)(img_1)

    dense_1 = Dropout(0.01)(Dense(64, activation=activations.relu, name="dense_1")(img_1))

    base_model = models.Model(inputs=inp, outputs=dense_1)
    opt = optimizers.Adam(0.001)

    base_model.compile(optimizer=opt, loss=losses.sparse_categorical_crossentropy, metrics=['acc'])
    model.summary()
    return base_model
model=get_base_model()

test_loss, test_acc = model.evaluate(Xtest, ytest, verbose=0) 
model.fit(X,y)
print('nTest accuracy:', test_acc)
 

Я получил ошибку :
Ввод 0 несовместим со слоем model_16: ожидаемая форма = (Нет, 335, 48), найденная форма = (Нет, 48)

Вот что вам нужно сделать.

1. В качестве входных данных вам не нужно указывать (355, 48) размер входных данных. Первый канал автоматически устанавливается на none . Установка его в (48,) значение позволит вам передавать 48 объектов для каждого образца.
2. Conv1D ожидает 2D-тензор, поэтому используйте layers.Reshape для изменения (48,) на (48,1)
3. Большое количество maxpools вызывает некоторые проблемы, поэтому я удалил некоторые блоки conv1d и обновил размеры ядра, чтобы у вас не заканчивались функции для свертки.

 from tensorflow.keras.layers import *
from tensorflow.keras import activations, losses, optimizers, models

def get_base_model():
    inp = Input(shape=(48,))
    x = Reshape((-1,1))(inp)
    img_1 = Convolution1D(16, kernel_size=3, activation=activations.relu, padding="valid")(x)
    img_1 = Convolution1D(16, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = MaxPool1D(pool_size=2)(img_1)
    img_1 = SpatialDropout1D(rate=0.01)(img_1)
    img_1 = Convolution1D(32, kernel_size=2, activation=activations.relu, padding="valid")(img_1)
    img_1 = Convolution1D(32, kernel_size=2, activation=activations.relu, padding="valid")(img_1)
    img_1 = MaxPool1D(pool_size=2)(img_1)
    img_1 = SpatialDropout1D(rate=0.01)(img_1)
    img_1 = Convolution1D(256, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = Convolution1D(256, kernel_size=3, activation=activations.relu, padding="valid")(img_1)
    img_1 = GlobalMaxPool1D()(img_1)
    img_1 = Dropout(rate=0.01)(img_1)

    dense_1 = Dropout(0.01)(Dense(64, activation=activations.relu, name="dense_1")(img_1))

    base_model = models.Model(inputs=inp, outputs=dense_1)
    opt = optimizers.Adam(0.001)

    base_model.compile(optimizer=opt, loss=losses.sparse_categorical_crossentropy, metrics=['acc'])
    model.summary()
    return base_model

model = get_base_model()

X = np.random.random((335, 48))
y = np.random.random((335,))

model.fit(X, y, epochs=4)
 

 Model: "functional_194"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_158 (InputLayer)       [(None, 48)]              0         
_________________________________________________________________
reshape_48 (Reshape)         (None, 48, 1)             0         
_________________________________________________________________
conv1d_57 (Conv1D)           (None, 46, 16)            64        
_________________________________________________________________
conv1d_58 (Conv1D)           (None, 44, 16)            784       
_________________________________________________________________
max_pooling1d_21 (MaxPooling (None, 22, 16)            0         
_________________________________________________________________
spatial_dropout1d_21 (Spatia (None, 22, 16)            0         
_________________________________________________________________
conv1d_59 (Conv1D)           (None, 21, 32)            1056      
_________________________________________________________________
conv1d_60 (Conv1D)           (None, 20, 32)            2080      
_________________________________________________________________
max_pooling1d_22 (MaxPooling (None, 10, 32)            0         
_________________________________________________________________
spatial_dropout1d_22 (Spatia (None, 10, 32)            0         
_________________________________________________________________
conv1d_61 (Conv1D)           (None, 8, 256)            24832     
_________________________________________________________________
conv1d_62 (Conv1D)           (None, 6, 256)            196864    
_________________________________________________________________
global_max_pooling1d_6 (Glob (None, 256)               0         
_________________________________________________________________
dropout_12 (Dropout)         (None, 256)               0         
_________________________________________________________________
dense_1 (Dense)              (None, 64)                16448     
_________________________________________________________________
dropout_13 (Dropout)         (None, 64)                0         
=================================================================
Total params: 242,128
Trainable params: 242,128
Non-trainable params: 0


Epoch 1/4
11/11 [==============================] - 0s 19ms/step - loss: 8.6069 - acc: 0.0000e 00
Epoch 2/4
11/11 [==============================] - 0s 9ms/step - loss: 4.1998 - acc: 0.0000e 00
Epoch 3/4
11/11 [==============================] - 0s 11ms/step - loss: 4.1589 - acc: 0.0000e 00
Epoch 4/4
11/11 [==============================] - 0s 10ms/step - loss: 4.1820 - acc: 0.0000e 00