![](/img/trans.png)
[英]Loading one dimensional data into a Dense layer in a sequential Keras model
[英]Mixing numerical and categorical data into keras sequential model with Dense layers
我在 Pandas 数据框中有一个训练集,我将这个数据框传递给带有df.values
的model.fit()
。 以下是有关 df 的一些信息:
df.values.shape
# (981, 5)
df.values[0]
# array([163, 0.6, 83, 0.52,
# array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
# 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
# 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
# 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
# 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
# 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
# 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
# 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
# 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
# 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
# 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
# 0, 0, 0, 0, 0, 0, 0])], dtype=object)
如您所见,df 中的行包含 5 列,其中 4 列包含数值(int 或 float),还有一个包含表示一些分类数据的热编码数组。 我正在创建我的 keras 模型,如下所示:
model = keras.Sequential([
keras.layers.Dense(1024, activation=tf.nn.relu, kernel_initializer=init_orth, bias_initializer=init_0),
keras.layers.Dense(512, activation=tf.nn.relu, kernel_initializer=init_orth, bias_initializer=init_0),
keras.layers.Dense(256, activation=tf.nn.relu, kernel_initializer=init_orth, bias_initializer=init_0),
keras.layers.Dense(128, activation=tf.nn.relu, kernel_initializer=init_orth, bias_initializer=init_0),
keras.layers.Dense(64, activation=tf.nn.relu, kernel_initializer=init_orth, bias_initializer=init_0),
keras.layers.Dense(1, activation=tf.nn.sigmoid)
])
opt = keras.optimizers.Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=None, decay=0.0, amsgrad=True)
model.compile(optimizer=opt,
loss='binary_crossentropy',
metrics=['accuracy'])
model.fit(df.values, df_labels.values, epochs=10, batch_size=32, verbose=0)
df_labels.values
只是一个 0 和 1 的一维数组。 所以我相信我最后确实需要一个 Dense(1) sigmoid 层,以及“binary_crossentropy”损失。
如果我只传递数字数据,这个模型效果很好。 但是一旦我引入热编码(分类数据),我就会得到这个错误:
ValueError Traceback (most recent call last)
<ipython-input-91-b5e6232b375f> in <module>
42 #trn_values = df_training_set.values[:,:,len(df_training_set.columns)]
43 #trn_cat = df_trn_wtid.values.reshape(-1, 1)
---> 44 model.fit(df_training_set.values, df_training_labels.values, epochs=10, batch_size=32, verbose=0)
45
46 #test_loss, test_acc = model.evaluate(df_test_set.values, df_test_labels.values)
~\Anaconda3\lib\site-packages\keras\engine\training.py in fit(self, x, y, batch_size, epochs, verbose, callbacks, validation_split, validation_data, shuffle, class_weight, sample_weight, initial_epoch, steps_per_epoch, validation_steps, **kwargs)
1037 initial_epoch=initial_epoch,
1038 steps_per_epoch=steps_per_epoch,
-> 1039 validation_steps=validation_steps)
1040
1041 def evaluate(self, x=None, y=None,
~\Anaconda3\lib\site-packages\keras\engine\training_arrays.py in fit_loop(model, f, ins, out_labels, batch_size, epochs, verbose, callbacks, val_f, val_ins, shuffle, callback_metrics, initial_epoch, steps_per_epoch, validation_steps)
197 ins_batch[i] = ins_batch[i].toarray()
198
--> 199 outs = f(ins_batch)
200 outs = to_list(outs)
201 for l, o in zip(out_labels, outs):
~\Anaconda3\lib\site-packages\keras\backend\tensorflow_backend.py in __call__(self, inputs)
2713 return self._legacy_call(inputs)
2714
-> 2715 return self._call(inputs)
2716 else:
2717 if py_any(is_tensor(x) for x in inputs):
~\Anaconda3\lib\site-packages\keras\backend\tensorflow_backend.py in _call(self, inputs)
2653 array_vals.append(
2654 np.asarray(value,
-> 2655 dtype=tf.as_dtype(tensor.dtype).as_numpy_dtype))
2656 if self.feed_dict:
2657 for key in sorted(self.feed_dict.keys()):
~\Anaconda3\lib\site-packages\numpy\core\numeric.py in asarray(a, dtype, order)
536
537 """
--> 538 return array(a, dtype, copy=False, order=order)
539
540
ValueError: setting an array element with a sequence.
请不要建议将 one_hot 数组中的每个值展开到它们自己的列中。 此示例是我的数据集的精简版本,其中包含 6-8 个分类列,其中一些 one_hots 是 5000+ 大小的数组。 所以这对我来说不是一个可行的解决方案。 我希望改进我的顺序模型(或彻底检查 keras 模型),以便处理分类数据和数值数据。
请记住,训练标签是 0/1 值的一维数组。 我需要预测一组结果的数值/分类训练集,我不能有一组来自数值数据的预测和一组来自分类数据的预测。
如果您仍在处理此问题,这是我从调用层时遇到的 tensorflow 异常中重用的内容(类型 CategoryEncoding)
只要您不从 1 更改时间步长,就应该没问题。我还包括了产生出色准确性的输出……因为我重用了标签的一个功能。 您的问题更多是关于如何将分类和数字输入模型。 你会从那里拿走它。
import tensorflow as tf
import pandas as pd
import numpy as np
# Simulate a data set of categorical and numerical values
# Configure simulation specifications: {feature: number of unique categories or None for numerical
theSimSpecs = {'Cat1': 54, 'Cat2': 2, 'Cat3': 4, 'Num1': None, 'Num2': None}
# batch size and timesteps
theBatchSz, theTimeSteps= 10, 1
# Creation of the dataset as pandas.DataFrame
theDFs = []
for theFeature, theUniques in theSimSpecs.items():
if theUniques is None:
theDF = pd.DataFrame(np.random.random(size=theBatchSz * theTimeSteps), columns=[theFeature])
else:
theDF = pd.DataFrame(np.random.randint(low=0, high=theUniques, size=theBatchSz * theTimeSteps),
columns=[theFeature]).astype('category')
theDFs.append(theDF)
theDF = pd.concat(theDFs, axis=1)
# code excerpt
# inventory of the categorical features' values ( None for the numerical)
theCatCodes = {theCol: (theDF[theCol].unique().tolist() if str(theDF[theCol].dtypes) == "category" else None)
for theCol in theDF.columns}
# Creation of the batched tensorflow.data.Dataset
theDS = tf.data.Dataset.from_tensor_slices(dict(theDF))
theDS = theDS.window(size=theTimeSteps, shift=1, stride=1, drop_remainder=True)
theDS = theDS.flat_map(lambda x: tf.data.Dataset.zip(x))
theDS = theDS.batch(batch_size=theBatchSz, drop_remainder=True)
# extracting one batch
theBatch = next(iter(theDS))
tf.print(theBatch)
# Creation of the components for the interface layer
theFeaturesInputs = {}
theFeaturesEncoded = {}
for theFeature, theCodes in theCatCodes.items():
if theCodes is None: # Pass-through for numerical features
theNumInput = tf.keras.layers.Input(shape=[], dtype=tf.float32, name=theFeature)
theFeaturesInputs[theFeature] = theNumInput
theFeatureExp = tf.expand_dims(input=theNumInput, axis=-1)
theFeaturesEncoded[theFeature] = theFeatureExp
else: # Process for categorical features
theCatInput = tf.keras.layers.Input(shape=[], dtype=tf.int64, name=theFeature)
theFeaturesInputs[theFeature] = theCatInput
theFeatureExp = tf.expand_dims(input=theCatInput, axis=-1)
theEncodingLayer = tf.keras.layers.CategoryEncoding(num_tokens=theSimSpecs[theFeature],
name=f"{theFeature}_enc",
output_mode="one_hot", sparse=False)
theFeaturesEncoded[theFeature] = theEncodingLayer(theFeatureExp)
### Below is what you'd be interested in
theStackedInputs = tf.concat(tf.nest.flatten(theFeaturesEncoded), axis=1)
theModel = tf.keras.Model(inputs=theFeaturesInputs, outputs=theStackedInputs)
theOutputs = theModel(theBatch)
tf.print(theOutputs[:5], summarize=-1)
x = tf.keras.layers.Dense(1024, activation=tf.nn.relu)(theStackedInputs)
x = tf.keras.layers.Dense(512, activation=tf.nn.relu)(x)
x = tf.keras.layers.Dense(256, activation=tf.nn.relu)(x)
x = tf.keras.layers.Dense(128, activation=tf.nn.relu)(x)
x = tf.keras.layers.Dense(64, activation=tf.nn.relu)(x)
theModelOutputs = tf.keras.layers.Dense(1, activation=tf.nn.sigmoid)(x)
theFullModel = tf.keras.Model(inputs=theFeaturesInputs, outputs=theModelOutputs)
opt = tf.keras.optimizers.Adam(learning_rate=0.001, beta_1=0.9, beta_2=0.999, epsilon=None, decay=0.0, amsgrad=True)
theFullModel.compile(optimizer=opt, loss='binary_crossentropy', metrics=['accuracy'])
# Yes! I was not super inspired and used one of the categorical columns as the label.
theFullModel.fit(x=theBatch, y=theBatch["Cat2"], epochs=10, verbose=1)
theOutput = theFullModel(theBatch)
tf.print(theOutput, summarize=-1)
# No big whoop, they match. I only showed how to write the way to get the categorical and numerical
# to be fed to the model in one go
tf.print(theBatch["Cat2"], summarize=-1)
{'Cat1': [10 49 43 ... 12 30 16], 'Cat2': [0 1 0 ... 0 0 1], 'Cat3': [0 0 3 ... 2 0 3], 'Num1': [0.61139996409794306 0.5939044614577218 0.59720125040942829 ... 0.60667441398817357 0.29892784677577522 0.18648129276910852], 'Num2': [0.10468088989623936 0.18483904850190647 0.053207335764727581 ... 0.065936326328562944 0.50927258084392213 0.14051443214527148]}
[[0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0.611399949 0.104680888]
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 1 0 0 0 0.593904436 0.184839055]
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0.597201228 0.0532073341]
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0.668277 0.012965099]
[0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 0 0.594983 0.969921947]]
Epoch 1/10
1/1 [==============================] - 1s 542ms/step - loss: 0.6840 - accuracy: 1.0000
Epoch 2/10
1/1 [==============================] - 0s 3ms/step - loss: 0.6190 - accuracy: 1.0000
Epoch 3/10
1/1 [==============================] - 0s 3ms/step - loss: 0.5387 - accuracy: 1.0000
Epoch 4/10
1/1 [==============================] - 0s 3ms/step - loss: 0.4398 - accuracy: 1.0000
Epoch 5/10
1/1 [==============================] - 0s 3ms/step - loss: 0.3275 - accuracy: 1.0000
Epoch 6/10
1/1 [==============================] - 0s 3ms/step - loss: 0.2174 - accuracy: 1.0000
Epoch 7/10
1/1 [==============================] - 0s 3ms/step - loss: 0.1264 - accuracy: 1.0000
Epoch 8/10
1/1 [==============================] - 0s 3ms/step - loss: 0.0643 - accuracy: 1.0000
Epoch 9/10
1/1 [==============================] - 0s 3ms/step - loss: 0.0280 - accuracy: 1.0000
Epoch 10/10
1/1 [==============================] - 0s 3ms/step - loss: 0.0104 - accuracy: 1.0000
[[0.000557512045]
[0.990028]
[0.000891149044]
[0.000809252262]
[0.000103618848]
[0.990051866]
[4.47019956e-05]
[0.000348180532]
[0.000630795956]
[0.990045428]]
[0 1 0 0 0 1 0 0 0 1]
Process finished with exit code 0
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.