This the code I'm working on:
这是我正在编写的代码:
from __future__ import print_function
from keras.models import Sequential
from keras.layers import Dense
from sklearn.cross_validation import train_test_split
import numpy
numpy.random.seed(7)
data_pixels=np.genfromtxt("pixels_dataset.csv", delimiter=',')
classes_dataset=np.genfromtxt("labels.csv",dtype=np.str , delimiter='\t')
x_train, x_test, y_train, y_test = train_test_split(data_pixels, classes_dataset, test_size=0.3
x_train
has a shape of (1708, 3072)
x_train有一个形状(1708,3072)
array([[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 1., 1., 1.],
[ 1., 1., 1., ..., 1., 1., 1.],
...,
[ 0., 0., 0., ..., 1., 1., 1.],
[ 1., 1., 1., ..., 1., 1., 1.],
[ 0., 0., 0., ..., 1., 1., 1.]])
y_train
has a shape of (1708,)
y_train有一个形状(1708,)
array(['7', 'f', '3', ..., '6', 'o', 'O'],
dtype='|S5')
the characters of y_train are
y_train的字符是。
: , : ; ! è à ä Aa..Zz 0-9
::;!e一个Aa . .Zz 0 - 9
model = Sequential()
model.add(Dense(12, input_dim=3072, activation='relu'))
model.add(Dense(8, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
l got error after executing the following :
在执行下列操作后,出现错误:
model.fit(x_train,y_train, epochs=150, batch_size=10)
the error is
错误的是
ValueError: could not convert string to float: A
l tried the following alternatives : 1)
我尝试了以下选项:1)
x_train=n.array(x_train)
y_train=n.array(y_train)
2)
2)
model.fit(x_train,str(y_train), epochs=150, batch_size=10)
But l got the same error Then l tried another alternative
但是我有同样的错误,我尝试了另一种选择。
from sklearn.preprocessing import LabelBinarizer
encoder = LabelBinarizer()
y_train = encoder.fit_transform(y_train)
then l get a new error which is
然后我得到一个新的错误。
ValueError: Error when checking model target: expected dense_21 to have shape (None, 1) but got array with shape (1708, 66)
1 个解决方案
#1
2
Change the following lines of code:
更改以下代码行:
model.add(Dense(66, activation='softmax'))
and:
和:
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
The problem lied in this that you wanted to predict a char
which was coded as one-hot
vector of length 66. In this case - you are setting your output to have desired length and you are using categorical_crossentropy
loss and softmax
activation.
问题在于,你想要预测一个字符,它被编码成一个长度为66的热矢量。在这种情况下,您将设置您的输出以获得所需的长度,并且您正在使用categorical_cross熵损失和softmax激活。
#1
2
Change the following lines of code:
更改以下代码行:
model.add(Dense(66, activation='softmax'))
and:
和:
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
The problem lied in this that you wanted to predict a char
which was coded as one-hot
vector of length 66. In this case - you are setting your output to have desired length and you are using categorical_crossentropy
loss and softmax
activation.
问题在于,你想要预测一个字符,它被编码成一个长度为66的热矢量。在这种情况下,您将设置您的输出以获得所需的长度,并且您正在使用categorical_cross熵损失和softmax激活。