Related
import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision
from torchvision import transforms, datasets
#neural network class
class Net(nn.Module):
#intialize class
def __init__(self):
super().__init__()
#feedforward neural network passes data from input layer to output layer
#fully connected layer with input shape of 28*28 pixels (flatten image to one row) and output feature is size 64. Linear means flat layer
self.fc1 = nn.Linear(28*28,64)
#feed in data from fc1 to fc2
self.fc2 = nn.Linear(64,64)
self.fc3 = nn.Linear(64,64)
#output layer has input 64 and output of size 10 to represent 10 classes in MNIST
self.fc4 = nn.Linear(64,10)
#forward pass through the data
def forward(self, x):
#relu is activation function and performs operation on input data
#input and output dimenson of relu are the same
x = F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
x = F.relu(self.fc3(x))
#softmax function gets probability distribution for each class adding up to 1 for output layer
x = F.log_softmax(self.fc4(x), dim=1)
return x
#declare a model
net = Net()
#print(net)
#passing in random data
x = torch.rand(28,28)
#resize to represent input shape (batch size, input x, input y)
x = x.view(-1,28,28)
#print(x)
#optmizer adjusts neural network based on error calculation
import torch.optim as optim
#net.parameters() means all the adjustable parts of the neural network, learning rate is amount of change (we don't want model to swerve based on one train case)
optimizer = optim.Adam(net.parameters(),lr=0.001)
#get datasets using torchvision.datasets transforms are application applied to data (transforms conversion to tensors)
train = torchvision.datasets.MNIST("", train=True, download=True,transform=transforms.Compose([transforms.ToTensor()]))
test = torchvision.datasets.MNIST("", train=False, download=True,transform=transforms.Compose([transforms.ToTensor()]))
#store in data loader, batch size is how many samples is passed through the model at once (in GPU memory), best batch size is between 8-64
#shuffling avoids feeding too much of one kind of image and leads to more generalization
trainset = torch.utils.data.DataLoader(train,batch_size=10,shuffle=True)
testset = torch.utils.data.DataLoader(test,batch_size=10,shuffle=True)
#full pass through data is epoch
EPOCHS = 3
for epoch in range(EPOCHS):
#data is a batch of data in the training set
for data in trainset:
#split into features and labels
features, labels = data
#print(features, labels)
#reset the gradient for next passes to avoid convoluting the results of multiple backpropogations
net.zero_grad()
#pass data into network (make sure input shape matches)
output = net(features.view(-1,28*28))
#compute error (output,expected)
loss = F.nll_loss(output,labels)
print("loss is the ", loss)
#backpropogate loss through trainiable parameters of model
loss.backward()
#adjust neural network
optimizer.step()
I am using Pytorch on Google colab. The error message says that the gradient isn't there.
I am unsure where the error stems from? I used this tutorial: https://www.youtube.com/watch?v=9j-_dOze4IM&list=PLQVvvaa0QuDdeMyHEYc0gxFpYwHY2Qfdh&index=4
Error message from Google colab
Error message
I am doing a slight modification of a standard neural network by defining a custom loss function. The custom loss function depends not only on y_true and y_pred, but also on the training data. I implemented it using the wrapping solution described here.
Specifically, I wanted to define a custom loss function that is the standard mse plus the mse between the input and the square of y_pred:
def custom_loss(x_true)
def loss(y_true, y_pred):
return K.mean(K.square(y_pred - y_true) + K.square(y_true - x_true))
return loss
Then I compile the model using
model_custom.compile(loss = custom_loss( x_true=training_data ), optimizer='adam')
fit the model using
model_custom.fit(training_data, training_label, epochs=100, batch_size = training_data.shape[0])
All of the above works fine, because the batch size is actually the number of all the training samples.
But if I set a different batch_size (e.g., 10) when I have 1000 training samples, there will be an error
Incompatible shapes: [1000] vs. [10].
It seems that Keras is able to automatically adjust the size of the inputs to its own loss function base on the batch size, but cannot do so for the custom loss function.
Do you know how to solve this issue?
Thank you!
==========================================================================
* Update: the batch size issue is solved, but another issue occurred
Thank you, Ori, for the suggestion of concatenating the input and output layers! It "worked", in the sense that the codes can run under any batch size. However, it seems that the result from training the new model is wrong... Below is a simplified version of the codes to demonstrate the problem:
import numpy as np
import scipy.io
import keras
from keras import backend as K
from keras.models import Model
from keras.layers import Input, Dense, Activation
from numpy.random import seed
from tensorflow import set_random_seed
def custom_loss(y_true, y_pred): # this is essentially the mean_square_error
mse = K.mean( K.square( y_pred[:,2] - y_true ) )
return mse
# set the seeds so that we get the same initialization across different trials
seed_numpy = 0
seed_tensorflow = 0
# generate data of x = [ y^3 y^2 ]
y = np.random.rand(5000+1000,1) * 2 # generate 5000 training and 1000 testing samples
x = np.concatenate( ( np.power(y, 3) , np.power(y, 2) ) , axis=1 )
training_data = x[0:5000:1,:]
training_label = y[0:5000:1]
testing_data = x[5000:6000:1,:]
testing_label = y[5000:6000:1]
# build the standard neural network with one hidden layer
seed(seed_numpy)
set_random_seed(seed_tensorflow)
input_standard = Input(shape=(2,)) # input
hidden_standard = Dense(10, activation='relu', input_shape=(2,))(input_standard) # hidden layer
output_standard = Dense(1, activation='linear')(hidden_standard) # output layer
model_standard = Model(inputs=[input_standard], outputs=[output_standard]) # build the model
model_standard.compile(loss='mean_squared_error', optimizer='adam') # compile the model
model_standard.fit(training_data, training_label, epochs=50, batch_size = 500) # train the model
testing_label_pred_standard = model_standard.predict(testing_data) # make prediction
# get the mean squared error
mse_standard = np.sum( np.power( testing_label_pred_standard - testing_label , 2 ) ) / 1000
# build the neural network with the custom loss
seed(seed_numpy)
set_random_seed(seed_tensorflow)
input_custom = Input(shape=(2,)) # input
hidden_custom = Dense(10, activation='relu', input_shape=(2,))(input_custom) # hidden layer
output_custom_temp = Dense(1, activation='linear')(hidden_custom) # output layer
output_custom = keras.layers.concatenate([input_custom, output_custom_temp])
model_custom = Model(inputs=[input_custom], outputs=[output_custom]) # build the model
model_custom.compile(loss = custom_loss, optimizer='adam') # compile the model
model_custom.fit(training_data, training_label, epochs=50, batch_size = 500) # train the model
testing_label_pred_custom = model_custom.predict(testing_data) # make prediction
# get the mean squared error
mse_custom = np.sum( np.power( testing_label_pred_custom[:,2:3:1] - testing_label , 2 ) ) / 1000
# compare the result
print( [ mse_standard , mse_custom ] )
Basically, I have a standard one-hidden-layer neural network, and a custom one-hidden-layer neural network whose output layer is concatenated with the input layer. For testing purpose, I did not use the concatenated input layer in the custom loss function, because I wanted to see if the custom network can reproduce the standard neural network. Since the custom loss function is equivalent to the standard 'mean_squared_error' loss, both networks should have the same training results (I also reset the random seeds to make sure that they have the same initialization).
However, the training results are very different. It seems that the concatenation makes the training process different? Any ideas?
Thank you again for all your help!
Final update: Ori's approach of concatenating input and output layers works, and is verified by using the generator. Thanks!!
The problem is that when compiling the model, you set x_true to be a static tensor, in the size of all the samples. While the input for keras loss functions are the y_true and y_pred, where each of them is of size [batch_size, :].
As I see it there are 2 options you can solve this, the first one is using a generator for creating the batches, in such a way that you will have control over which indices are evaluated each time, and at the loss function you could slice the x_true tensor to fit the samples being evaluated:
def custom_loss(x_true)
def loss(y_true, y_pred):
x_true_samples = relevant_samples(x_true)
return K.mean(K.square(y_pred - y_true) + K.square(y_true - x_true_samples))
return loss
This solution can be complicated, what I would suggest is a simpler workaround -
Concatenate the input layer with the output layer, such that your new output will be of the form original_output , input.
Now you can use a new modified loss function:
def loss(y_true, y_pred):
return K.mean(K.square(y_pred[:,:output_shape] - y_true[:,:output_shape]) +
K.square(y_true[:,:output_shape] - y_pred[:,outputshape:))
Now your new loss function will take in account both the input data, and the prediction.
Edit:
Note that while you set the seed, your models are not exactly the same, and as you did not use a generator, you let keras choose the batches, and for different models he might pick different samples.
As your model does not converge, different samples can lead to different results.
I added a generator to your code, to verify the samples we pick for training, now you can see both results are the same:
def custom_loss(y_true, y_pred): # this is essentially the mean_square_error
mse = keras.losses.mean_squared_error(y_true, y_pred[:,2])
return mse
def generator(x, y, batch_size):
curIndex = 0
batch_x = np.zeros((batch_size,2))
batch_y = np.zeros((batch_size,1))
while True:
for i in range(batch_size):
batch_x[i] = x[curIndex,:]
batch_y[i] = y[curIndex,:]
i += 1;
if i == 5000:
i = 0
yield batch_x, batch_y
# set the seeds so that we get the same initialization across different trials
seed_numpy = 0
seed_tensorflow = 0
# generate data of x = [ y^3 y^2 ]
y = np.random.rand(5000+1000,1) * 2 # generate 5000 training and 1000 testing samples
x = np.concatenate( ( np.power(y, 3) , np.power(y, 2) ) , axis=1 )
training_data = x[0:5000:1,:]
training_label = y[0:5000:1]
testing_data = x[5000:6000:1,:]
testing_label = y[5000:6000:1]
batch_size = 32
# build the standard neural network with one hidden layer
seed(seed_numpy)
set_random_seed(seed_tensorflow)
input_standard = Input(shape=(2,)) # input
hidden_standard = Dense(10, activation='relu', input_shape=(2,))(input_standard) # hidden layer
output_standard = Dense(1, activation='linear')(hidden_standard) # output layer
model_standard = Model(inputs=[input_standard], outputs=[output_standard]) # build the model
model_standard.compile(loss='mse', optimizer='adam') # compile the model
#model_standard.fit(training_data, training_label, epochs=50, batch_size = 10) # train the model
model_standard.fit_generator(generator(training_data,training_label,batch_size), steps_per_epoch= 32, epochs= 100)
testing_label_pred_standard = model_standard.predict(testing_data) # make prediction
# get the mean squared error
mse_standard = np.sum( np.power( testing_label_pred_standard - testing_label , 2 ) ) / 1000
# build the neural network with the custom loss
seed(seed_numpy)
set_random_seed(seed_tensorflow)
input_custom = Input(shape=(2,)) # input
hidden_custom = Dense(10, activation='relu', input_shape=(2,))(input_custom) # hidden layer
output_custom_temp = Dense(1, activation='linear')(hidden_custom) # output layer
output_custom = keras.layers.concatenate([input_custom, output_custom_temp])
model_custom = Model(inputs=input_custom, outputs=output_custom) # build the model
model_custom.compile(loss = custom_loss, optimizer='adam') # compile the model
#model_custom.fit(training_data, training_label, epochs=50, batch_size = 10) # train the model
model_custom.fit_generator(generator(training_data,training_label,batch_size), steps_per_epoch= 32, epochs= 100)
testing_label_pred_custom = model_custom.predict(testing_data)
# get the mean squared error
mse_custom = np.sum( np.power( testing_label_pred_custom[:,2:3:1] - testing_label , 2 ) ) / 1000
# compare the result
print( [ mse_standard , mse_custom ] )
I have used Keras and TensorFlow to classify the Fashion MNIST following this tutorial .
It uses the AdamOptimizer to find the value for model parameters that minimize the loss function of the network. The input for the network is a 2-D tensor with shape [28, 28], and output is a 1-D tensor with shape [10] which is the result of a softmax function.
Once the network has been trained, I want to use the optimizer for another task: find an input that maximizes one of the elements of the output tensor. How can this be done? Is it possible to do so using Keras or one have to use a lower level API?
Since the input is not unique for a given output, it would be even better if we could impose some constraints on the values the input can take.
The trained model has the following format
model = keras.Sequential([
keras.layers.Flatten(input_shape=(28, 28)),
keras.layers.Dense(128, activation=tf.nn.relu),
keras.layers.Dense(10, activation=tf.nn.softmax)
])
I feel you would want to backprop with respect to the input freezing all the weights to your model. What you could do is:
Add a dense layer after the input layer with the same dimensions as input and set it as trainable
Freeze all the other layers of your model. (except the one you added)
As an input, feed an identity matrix and train your model based on whatever output you desire.
This article and this post might be able to help you if you want to backprop based on the input instead. It's a bit like what you are aiming for but you can get the intuition.
It would be very similar to the way that filters of a Convolutional Network is visualized: we would do gradient ascent optimization in input space to maximize the response of a particular filter.
Here is how to do it: after training is finished, first we need to specify the output and define a loss function that we want to maximize:
from keras import backend as K
output_class = 0 # the index of the output class we want to maximize
output = model.layers[-1].output
loss = K.mean(output[:,output_class]) # get the average activation of our desired class over the batch
Next, we need to take the gradient of the loss we have defined above with respect to the input layer:
grads = K.gradients(loss, model.input)[0] # the output of `gradients` is a list, just take the first (and only) element
grads = K.l2_normalize(grads) # normalize the gradients to help having an smooth optimization process
Next, we need to define a backend function that takes the initial input image and gives the values of loss and gradients as outputs, so that we can use it in the next step to implement the optimization process:
func = K.function([model.input], [loss, grads])
Finally, we implement the gradient ascent optimization process:
import numpy as np
input_img = np.random.random((1, 28, 28)) # define an initial random image
lr = 1. # learning rate used for gradient updates
max_iter = 50 # number of gradient updates iterations
for i in range(max_iter):
loss_val, grads_val = func([input_img])
input_img += grads_val * lr # update the image based on gradients
Note that, after this process is finished, to display the image you may need to make sure that all the values in the image are in the range [0, 255] (or [0,1]).
After the hints Saket Kumar Singh gave in his answer, I wrote the following that seems to solve the question.
I create two custom layers. Maybe Keras offers already some classes that are equivalent to them.
The first on is a trainable input:
class MyInputLayer(keras.layers.Layer):
def __init__(self, output_dim, **kwargs):
self.output_dim = output_dim
super(MyInputLayer, self).__init__(**kwargs)
def build(self, input_shape):
self.kernel = self.add_weight(name='kernel',
shape=self.output_dim,
initializer='uniform',
trainable=True)
super(MyInputLayer, self).build(input_shape)
def call(self, x):
return self.kernel
def compute_output_shape(self, input_shape):
return self.output_dim
The second one gets the probability of the label of interest:
class MySelectionLayer(keras.layers.Layer):
def __init__(self, position, **kwargs):
self.position = position
self.output_dim = 1
super(MySelectionLayer, self).__init__(**kwargs)
def build(self, input_shape):
super(MySelectionLayer, self).build(input_shape)
def call(self, x):
mask = np.array([False]*x.shape[-1])
mask[self.position] = True
return tf.boolean_mask(x, mask,axis=1)
def compute_output_shape(self, input_shape):
return self.output_dim
I used them in this way:
# Build the model
layer_flatten = keras.layers.Flatten(input_shape=(28, 28))
layerDense1 = keras.layers.Dense(128, activation=tf.nn.relu)
layerDense2 = keras.layers.Dense(10, activation=tf.nn.softmax)
model = keras.Sequential([
layer_flatten,
layerDense1,
layerDense2
])
# Compile the model
model.compile(optimizer=tf.train.AdamOptimizer(),
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
# Train the model
# ...
# Freeze the model
layerDense1.trainable = False
layerDense2.trainable = False
# Build another model
class_index = 7
layerInput = MyInputLayer((1,784))
layerSelection = MySelectionLayer(class_index)
model_extended = keras.Sequential([
layerInput,
layerDense1,
layerDense2,
layerSelection
])
# Compile it
model_extended.compile(optimizer=tf.train.AdamOptimizer(),
loss='mean_absolute_error')
# Train it
dummyInput = np.ones((1,1))
target = np.ones((1,1))
model_extended.fit(dummyInput, target,epochs=300)
# Retrieve the weights of layerInput
layerInput.get_weights()[0]
Interesting. Maybe a solution would be to feed all your data to the network and for each sample save the output_layer after softmax.
This way, for 3 classes, where you want to find the best input for class 1, you are looking for outputs where the first component is high. For example: [1 0 0]
Indeed the output means the probability, or the confidence of the network, for the sample being one of the classes.
Funny coincident I was just working on the same "problem". I'm interested in the direction of adversarial training etc. What I did was to insert a LocallyConnected2D Layer after the input and then train with data which is all one and has as targets the class of interest.
As model I use
batch_size = 64
num_classes = 10
epochs = 20
input_shape = (28, 28, 1)
inp = tf.keras.layers.Input(shape=input_shape)
conv1 = tf.keras.layers.Conv2D(32, kernel_size=(3, 3),activation='relu',kernel_initializer='he_normal')(inp)
pool1 = tf.keras.layers.MaxPool2D((2, 2))(conv1)
drop1 = tf.keras.layers.Dropout(0.20)(pool1)
flat = tf.keras.layers.Flatten()(drop1)
fc1 = tf.keras.layers.Dense(128, activation='relu')(flat)
norm1 = tf.keras.layers.BatchNormalization()(fc1)
dropfc1 = tf.keras.layers.Dropout(0.25)(norm1)
out = tf.keras.layers.Dense(num_classes, activation='softmax')(dropfc1)
model = tf.keras.models.Model(inputs = inp , outputs = out)
model.compile(loss=tf.keras.losses.categorical_crossentropy,
optimizer=tf.keras.optimizers.RMSprop(),
metrics=['accuracy'])
model.summary()
after training I insert the new layer
def insert_intermediate_layer_in_keras(model,position, before_layer_id):
layers = [l for l in model.layers]
if(before_layer_id==0) :
x = new_layer
else:
x = layers[0].output
for i in range(1, len(layers)):
if i == before_layer_id:
x = new_layer(x)
x = layers[i](x)
else:
x = layers[i](x)
new_model = tf.keras.models.Model(inputs=layers[0].input, outputs=x)
return new_model
def fix_model(model):
for l in model.layers:
l.trainable=False
fix_model(model)
new_layer = tf.keras.layers.LocallyConnected2D(1, kernel_size=(1, 1),
activation='linear',
kernel_initializer='he_normal',
use_bias=False)
new_model = insert_intermediate_layer_in_keras(model,new_layer,1)
new_model.compile(loss=tf.keras.losses.categorical_crossentropy,
optimizer=tf.keras.optimizers.RMSprop(),
metrics=['accuracy'])
and finally rerun training with my fake data.
X_fake = np.ones((60000,28,28,1))
print(Y_test.shape)
y_fake = np.ones((60000))
Y_fake = tf.keras.utils.to_categorical(y_fake, num_classes)
new_model.fit(X_fake, Y_fake, epochs=100)
weights = new_layer.get_weights()[0]
imshow(weights.reshape(28,28))
plt.show()
Results are not yet satisfying but I'm confident of the approach and guess I need to play around with the optimiser.
I was going through this tutorial. I have a question about the following class code:
class RNN(nn.Module):
def __init__(self, input_size, hidden_size, output_size):
super(RNN, self).__init__()
self.input_size = input_size
self.hidden_size = hidden_size
self.output_size = output_size
self.i2h = nn.Linear(input_size + hidden_size, hidden_size)
self.i2o = nn.Linear(input_size + hidden_size, output_size)
self.softmax = nn.LogSoftmax()
def forward(self, input, hidden):
combined = torch.cat((input, hidden), 1)
hidden = self.i2h(combined)
output = self.i2o(combined)
output = self.softmax(output)
return output, hidden
def init_hidden(self):
return Variable(torch.zeros(1, self.hidden_size))
This code was taken from Here. There it was mentioned that
Since the state of the network is held in the graph and not in the layers, you can simply create an nn.Linear and reuse it over and over again for the recurrence.
What I don't understand is, how can one just increase input feature size in nn.Linear and say it is a RNN. What am I missing here?
The network is recurrent, because you evaluate multiple timesteps in the example.
The following code is also taken from the pytorch tutorial you linked to.
loss_fn = nn.MSELoss()
batch_size = 10
TIMESTEPS = 5
# Create some fake data
batch = torch.randn(batch_size, 50)
hidden = torch.zeros(batch_size, 20)
target = torch.zeros(batch_size, 10)
loss = 0
for t in range(TIMESTEPS):
# yes! you can reuse the same network several times,
# sum up the losses, and call backward!
hidden, output = rnn(batch, hidden)
loss += loss_fn(output, target)
loss.backward()
So the network itself is not recurrent, but in this loop you use it as a recurrent network by feeding the hidden state of the previous forward step together with your batch-input multiple times.
You could also use it non-recurrent by just backpropagating the loss in every step and ignoring the hidden state.
Since the state of the network is held in the graph and not in the layers, you can simply create an nn.Linear and reuse it over and over again for the recurrence.
This means, that the information to compute the gradient is not held in the model itself, so you can append multiple evaluations of the module to the graph and then backpropagate through the full graph.
This is described in the previous paragraphs of the tutorial.
I am trying to detect micro-events in a long time series. For this purpose, I will train a LSTM network.
Data. Input for each time sample is 11 different features somewhat normalized to fit 0-1. Output will be either one of two classes.
Batching. Due to huge class imbalance I have extracted the data in batches of each 60 time samples, of which at least 5 will always be class 1, and the rest class to. In this way the class imbalance is reduced from 150:1 to around 12:1 I have then randomized the order of all my batches.
Model. I am attempting to train an LSTM, with initial configuration of 3 different cells with 5 delay steps. I expect the micro events to arrive in sequences of at least 3 time steps.
Problem: When I try to train the network it will quickly converge towards saying that EVERYTHING belongs to the majority class. When I implement a weighted loss function, at some certain threshold it will change to saying that EVERYTHING belongs to the minority class. I suspect (without being expert) that there is no learning in my LSTM cells, or that my configuration is off?
Below is the code for my implementation. I am hoping that someone can tell me
Is my implementation correct?
What other reasons could there be for such behaviour?
ar_model.py
import numpy as np
import tensorflow as tf
from tensorflow.models.rnn import rnn
import ar_config
config = ar_config.get_config()
class ARModel(object):
def __init__(self, is_training=False, config=None):
# Config
if config is None:
config = ar_config.get_config()
# Placeholders
self._features = tf.placeholder(tf.float32, [None, config.num_features], name='ModelInput')
self._targets = tf.placeholder(tf.float32, [None, config.num_classes], name='ModelOutput')
# Hidden layer
with tf.variable_scope('lstm') as scope:
lstm_cell = tf.nn.rnn_cell.BasicLSTMCell(config.num_hidden, forget_bias=0.0)
cell = tf.nn.rnn_cell.MultiRNNCell([lstm_cell] * config.num_delays)
self._initial_state = cell.zero_state(config.batch_size, dtype=tf.float32)
outputs, state = rnn.rnn(cell, [self._features], dtype=tf.float32)
# Output layer
output = outputs[-1]
softmax_w = tf.get_variable('softmax_w', [config.num_hidden, config.num_classes], tf.float32)
softmax_b = tf.get_variable('softmax_b', [config.num_classes], tf.float32)
logits = tf.matmul(output, softmax_w) + softmax_b
# Evaluate
ratio = (60.00 / 5.00)
class_weights = tf.constant([ratio, 1 - ratio])
weighted_logits = tf.mul(logits, class_weights)
loss = tf.nn.softmax_cross_entropy_with_logits(weighted_logits, self._targets)
self._cost = cost = tf.reduce_mean(loss)
self._predict = tf.argmax(tf.nn.softmax(logits), 1)
self._correct = tf.equal(tf.argmax(logits, 1), tf.argmax(self._targets, 1))
self._accuracy = tf.reduce_mean(tf.cast(self._correct, tf.float32))
self._final_state = state
if not is_training:
return
# Optimize
optimizer = tf.train.AdamOptimizer()
self._train_op = optimizer.minimize(cost)
#property
def features(self):
return self._features
#property
def targets(self):
return self._targets
#property
def cost(self):
return self._cost
#property
def accuracy(self):
return self._accuracy
#property
def train_op(self):
return self._train_op
#property
def predict(self):
return self._predict
#property
def initial_state(self):
return self._initial_state
#property
def final_state(self):
return self._final_state
ar_train.py
import os
from datetime import datetime
import numpy as np
import tensorflow as tf
from tensorflow.python.platform import gfile
import ar_network
import ar_config
import ar_reader
config = ar_config.get_config()
def main(argv=None):
if gfile.Exists(config.train_dir):
gfile.DeleteRecursively(config.train_dir)
gfile.MakeDirs(config.train_dir)
train()
def train():
train_data = ar_reader.ArousalData(config.train_data, num_steps=config.max_steps)
test_data = ar_reader.ArousalData(config.test_data, num_steps=config.max_steps)
with tf.Graph().as_default(), tf.Session() as session, tf.device('/cpu:0'):
initializer = tf.random_uniform_initializer(minval=-0.1, maxval=0.1)
with tf.variable_scope('model', reuse=False, initializer=initializer):
m = ar_network.ARModel(is_training=True)
s = tf.train.Saver(tf.all_variables())
tf.initialize_all_variables().run()
for batch_input, batch_target in train_data:
step = train_data.iter_steps
dict = {
m.features: batch_input,
m.targets: batch_target
}
session.run(m.train_op, feed_dict=dict)
state, cost, accuracy = session.run([m.final_state, m.cost, m.accuracy], feed_dict=dict)
if not step % 10:
test_input, test_target = test_data.next()
test_accuracy = session.run(m.accuracy, feed_dict={
m.features: test_input,
m.targets: test_target
})
now = datetime.now().time()
print ('%s | Iter %4d | Loss= %.5f | Train= %.5f | Test= %.3f' % (now, step, cost, accuracy, test_accuracy))
if not step % 1000:
destination = os.path.join(config.train_dir, 'ar_model.ckpt')
s.save(session, destination)
if __name__ == '__main__':
tf.app.run()
ar_config.py
class Config(object):
# Directories
train_dir = '...'
ckpt_dir = '...'
train_data = '...'
test_data = '...'
# Data
num_features = 13
num_classes = 2
batch_size = 60
# Model
num_hidden = 3
num_delays = 5
# Training
max_steps = 100000
def get_config():
return Config()
UPDATED ARCHITECTURE:
# Placeholders
self._features = tf.placeholder(tf.float32, [None, config.num_features, config.num_delays], name='ModelInput')
self._targets = tf.placeholder(tf.float32, [None, config.num_output], name='ModelOutput')
# Weights
weights = {
'hidden': tf.get_variable('w_hidden', [config.num_features, config.num_hidden], tf.float32),
'out': tf.get_variable('w_out', [config.num_hidden, config.num_classes], tf.float32)
}
biases = {
'hidden': tf.get_variable('b_hidden', [config.num_hidden], tf.float32),
'out': tf.get_variable('b_out', [config.num_classes], tf.float32)
}
#Layer in
with tf.variable_scope('input_hidden') as scope:
inputs = self._features
inputs = tf.transpose(inputs, perm=[2, 0, 1]) # (BatchSize,NumFeatures,TimeSteps) -> (TimeSteps,BatchSize,NumFeatures)
inputs = tf.reshape(inputs, shape=[-1, config.num_features]) # (TimeSteps,BatchSize,NumFeatures -> (TimeSteps*BatchSize,NumFeatures)
inputs = tf.add(tf.matmul(inputs, weights['hidden']), biases['hidden'])
#Layer hidden
with tf.variable_scope('hidden_hidden') as scope:
inputs = tf.split(0, config.num_delays, inputs) # -> n_steps * (batchsize, features)
cell = tf.nn.rnn_cell.BasicLSTMCell(config.num_hidden, forget_bias=0.0)
self._initial_state = cell.zero_state(config.batch_size, dtype=tf.float32)
outputs, state = rnn.rnn(cell, inputs, dtype=tf.float32)
#Layer out
with tf.variable_scope('hidden_output') as scope:
output = outputs[-1]
logits = tf.add(tf.matmul(output, weights['out']), biases['out'])
Odd elements
Weighted loss
I am not sure your "weighted loss" does what you want it to do:
ratio = (60.00 / 5.00)
class_weights = tf.constant([ratio, 1 - ratio])
weighted_logits = tf.mul(logits, class_weights)
this is applied before calculating the loss function (further I think you wanted an element-wise multiplication as well? also your ratio is above 1 which makes the second part negative?) so it forces your predictions to behave in a certain way before applying the softmax.
If you want weighted loss you should apply this after
loss = tf.nn.softmax_cross_entropy_with_logits(weighted_logits, self._targets)
with some element-wise multiplication of your weights.
loss = loss * weights
Where your weights have a shape like [2,]
However, I would not recommend you to use weighted losses. Perhaps try increasing the ratio even further than 1:6.
Architecture
As far as I can read, you are using 5 stacked LSTMs with 3 hidden units per layer?
Try removing the multi rnn and just use a single LSTM/GRU (maybe even just a vanilla RNN) and jack the hidden units up to ~100-1000.
Debugging
Often when you are facing problems with an odd behaving network, it can be a good idea to:
Print everything
Literally print the shapes and values of every tensor in your model, use sess to fetch it and then print it. Your input data, the first hidden representation, your predictions, your losses etc.
You can also use tensorflows tf.Print() x_tensor = tf.Print(x_tensor, [tf.shape(x_tensor)])
Use tensorboard
Using tensorboard summaries on your gradients, accuracy metrics and histograms will reveal patterns in your data that might explain certain behavior, such as what lead to exploding weights. Like maybe your forget bias goes to infinity or your not tracking gradient through a certain layer etc.
Other questions
How large is your dataset?
How long are your sequences?
Are the 13 features categorical or continuous? You should not normalize categorical variables or represent them as integers, instead you should use one-hot encoding.
Gunnar has already made lots of good suggestions. A few more small things worth paying attention to in general for this sort of architecture:
Try tweaking the Adam learning rate. You should determine the proper learning rate by cross-validation; as a rough start, you could just check whether a smaller learning rate saves your model from crashing on the training data.
You should definitely use more hidden units. It's cheap to try larger networks when you first start out on a dataset. Go as large as necessary to avoid the underfitting you've observed. Later you can regularize / pare down the network after you get it to learn something useful.
Concretely, how long are the sequences you are passing into the network? You say you have a 30k-long time sequence.. I assume you are passing in subsections / samples of this sequence?