The training data (including both training and validation set) has about 80 million samples and each sample has 200 dense floating points. There are 6 labeled classe and they are imbalanced.
In the common-used ML libs (e.g., libsvm, scikit-learn, Spark MLlib, random forest, XGBoost or else), which should I use? Regarding the hardware configuration, the machine has 24 CPU cores and 250 Gb memory.
I would recommend using scikit-learn's SGDClassifier as it is online so you can load your training data in chunks (mini-batches) into memory and train the classifier gradually so you won't need to load all the data into memory.
It is highly parallel and easy to use.
You can set the warm_start argument to True and call fit multiple times with each chunk of X, y loaded into memory or the better option you can use the partial_fit method.
clf = SGDClassifier(loss='hinge', alpha=1e-4, penalty='l2', l1_ratio=0.9, learning_rate='optimal', n_iter=10, shuffle=False, n_jobs=10, fit_intercept=True)
# len(classes) = n_classes
all_classes = np.array(set_of_all_classes)
while True:
#load a minibatch from disk into memory
X, y = load_next_chunk()
clf.partial_fit(X, y, all_classes)
X_test, y_test = load_test_data()
y_pred = clf.predict(X_test)
Related
I have a dataset of 1mln * 44.
When I load data into KMeans - I get OOM (very low RAM)
I looked at MiniBatch K which has a partial_fit method.
I understand correctly that this code will give approximately the same result?
kmeans = MiniBatchKMeans(n_clusters=4, random_state=0, batch_size=100_000, n_init="auto")
for read_csv(csv_path, csv_names, csv_delimiter, chunksize = 100_000):
model.partial_fit(X)
Why can't I stream data into KMeans in sklearn? Using centroids of previous calculations for next calculations?
I have multi class labels and want to compute the accuracy of my model.
I am kind of confused on which sklearn function I need to use.
As far as I understood the below code is only used for the binary classification.
# dividing X, y into train and test data
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25,random_state = 0)
# training a linear SVM classifier
from sklearn.svm import SVC
svm_model_linear = SVC(kernel = 'linear', C = 1).fit(X_train, y_train)
svm_predictions = svm_model_linear.predict(X_test)
# model accuracy for X_test
accuracy = svm_model_linear.score(X_test, y_test)
print accuracy
and as I understood from the link:
Which decision_function_shape for sklearn.svm.SVC when using OneVsRestClassifier?
for multiclass classification I should use OneVsRestClassifier with decision_function_shape (with ovr or ovo and check which one works better)
svm_model_linear = OneVsRestClassifier(SVC(kernel = 'linear',C = 1, decision_function_shape = 'ovr')).fit(X_train, y_train)
The main problem is that the time of predicting the labels does matter to me but it takes about 1 minute to run the classifier and predict the data (also this time is added to the feature reduction such as PCA which also takes sometime)? any suggestions to reduce the time for svm multiclassifer?
There are multiple things to consider here:
1) You see, OneVsRestClassifier will separate out all labels and train multiple svm objects (one for each label) on the given data. So each time, only binary data will be supplied to single svm object.
2) SVC internally uses libsvm and liblinear, which have a 'OvO' strategy for multi-class or multi-label output. But this point will be of no use because of point 1. libsvm will only get binary data.
Even if it did, it doesnt take into account the 'decision_function_shape'. So it does not matter if you provide decision_function_shape = 'ovr' or decision_function_shape = 'ovr'.
So it seems that you are looking at the problem wrong. decision_function_shape should not affect the speed. Try standardizing your data before fitting. SVMs work well with standardized data.
When wrapping models with the ovr or ovc classifiers, you could set the n_jobs parameters to make them run faster, e.g. sklearn.multiclass.OneVsOneClassifier(estimator, n_jobs=-1) or sklearn.multiclass.OneVsRestClassifier(estimator, n_jobs=-1).
Although each single SVM classifier in sklearn could only use one CPU core at a time, the ensemble multi class classifier could fit multiple models at the same time by setting n_jobs.
I am trying to predict the hygrothermal response of a wall, given the interior and exterior climate. Based on literature research, I believe this should be possible with RNN but I have not been able to get good accuracy.
The dataset has 12 input features (time-series of exterior and interior climate data) and 10 output features (time-series of hygrothermal response), both containing hourly values for 10 years. This data was created with hygrothermal simulation software, there is no missing data.
Dataset features:
Dataset targets:
Unlike most time-series prediction problems, I want to predict the response for the full length of the input features time-series at each time-step, rather than the subsequent values of a time-series (eg financial time-series prediction). I have not been able to find similar prediction problems (in similar or other fields), so if you know of one, references are very welcome.
I think this should be possible with RNN, so I am currently using LSTM from Keras. Before training, I preprocess my data the following way:
Discard first year of data, as the first time steps of the hygrothermal response of the wall is influenced by the initial temperature and relative humidity.
Split into training and testing set. Training set contains the first 8 years of data, the test set contains the remaining 2 years.
Normalise training set (zero mean, unit variance) using StandardScaler from Sklearn. Normalise test set analogously using mean an variance from training set.
This results in: X_train.shape = (1, 61320, 12), y_train.shape = (1, 61320, 10), X_test.shape = (1, 17520, 12), y_test.shape = (1, 17520, 10)
As these are long time-series, I use stateful LSTM and cut the time-series as explained here, using the stateful_cut() function. I only have 1 sample, so batch_size is 1. For T_after_cut I have tried 24 and 120 (24*5); 24 appears to give better results. This results in X_train.shape = (2555, 24, 12), y_train.shape = (2555, 24, 10), X_test.shape = (730, 24, 12), y_test.shape = (730, 24, 10).
Next, I build and train the LSTM model as follows:
model = Sequential()
model.add(LSTM(128,
batch_input_shape=(batch_size,T_after_cut,features),
return_sequences=True,
stateful=True,
))
model.addTimeDistributed(Dense(targets)))
model.compile(loss='mean_squared_error', optimizer=Adam())
model.fit(X_train, y_train, epochs=100, batch_size=batch=batch_size, verbose=2, shuffle=False)
Unfortunately, I don't get accurate prediction results; not even for the training set, thus the model has high bias.
The prediction results of the LSTM model for all targets
How can I improve my model? I have already tried the following:
Not discarding the first year of the dataset -> no significant difference
Differentiating the input features time-series (subtract previous value from current value) -> slightly worse results
Up to four stacked LSTM layers, all with the same hyperparameters -> no significant difference in results but longer training time
Dropout layer after LSTM layer (though this is usually used to reduce variance and my model has high bias) -> slightly better results, but difference might not be statistically significant
Am I doing something wrong with the stateful LSTM? Do I need to try different RNN models? Should I preprocess the data differently?
Furthermore, training is very slow: about 4 hours for the model above. Hence I am reluctant to do an extensive hyperparameter gridsearch...
In the end, I managed to solve this the following way:
Using more samples to train instead of only 1 (I used 18 samples to train and 6 to test)
Keep the first year of data, as the output time-series for all samples have the same 'starting point' and the model needs this information to learn
Standardise both input and output features (zero mean, unit variance). I found this improved prediction accuracy and training speed
Use stateful LSTM as described here, but add reset states after epoch (see below for code). I used batch_size = 6 and T_after_cut = 1460. If T_after_cut is longer, training is slower; if T_after_cut is shorter, accuracy decreases slightly. If more samples are available, I think using a larger batch_size will be faster.
use CuDNNLSTM instead of LSTM, this speed up the training time x4!
I found that more units resulted in higher accuracy and faster convergence (shorter training time). Also I found that the GRU is as accurate as the LSTM tough converged faster for the same number of units.
Monitor validation loss during training and use early stopping
The LSTM model is build and trained as follows:
def define_reset_states_batch(nb_cuts):
class ResetStatesCallback(Callback):
def __init__(self):
self.counter = 0
def on_batch_begin(self, batch, logs={}):
# reset states when nb_cuts batches are completed
if self.counter % nb_cuts == 0:
self.model.reset_states()
self.counter += 1
def on_epoch_end(self, epoch, logs={}):
# reset states after each epoch
self.model.reset_states()
return(ResetStatesCallback)
model = Sequential()
model.add(layers.CuDNNLSTM(256, batch_input_shape=(batch_size,T_after_cut ,features),
return_sequences=True,
stateful=True))
model.add(layers.TimeDistributed(layers.Dense(targets, activation='linear')))
optimizer = RMSprop(lr=0.002)
model.compile(loss='mean_squared_error', optimizer=optimizer)
earlyStopping = EarlyStopping(monitor='val_loss', min_delta=0.005, patience=15, verbose=1, mode='auto')
ResetStatesCallback = define_reset_states_batch(nb_cuts)
model.fit(X_dev, y_dev, epochs=n_epochs, batch_size=n_batch, verbose=1, shuffle=False, validation_data=(X_eval,y_eval), callbacks=[ResetStatesCallback(), earlyStopping])
This gave me very statisfying accuracy (R2 over 0.98):
This figure shows the temperature (left) and relative humidity (right) in the wall over 2 years (data not used in training), prediction in red and true output in black. The residuals show that the error is very small and that the LSTM learns to capture the long-term dependencies to predict the relative humidity.
I'm a newbie to machine learning and this is one of the first real-world ML tasks challenged.
Some experimental data contains 512 independent boolean features and a boolean result.
There are about 1e6 real experiment records in the provided data set.
In a classic XOR example all 4 out of 4 possible states are required to train NN. In my case its only 2^(10-512) = 2^-505 which is close to zero.
I have no more information about the data nature, just these (512 + 1) * 1e6 bits.
Tried NN with 1 hidden layer on available data. Output of the trained NN on the samples even from the training set are always close to 0, not a single close to "1". Played with weights initialization, gradient descent learning rate.
My code utilizing TensorFlow 1.3, Python 3. Model excerpt:
with tf.name_scope("Layer1"):
#W1 = tf.Variable(tf.random_uniform([512, innerN], minval=-2/512, maxval=2/512), name="Weights_1")
W1 = tf.Variable(tf.zeros([512, innerN]), name="Weights_1")
b1 = tf.Variable(tf.zeros([1]), name="Bias_1")
Out1 = tf.sigmoid( tf.matmul(x, W1) + b1)
with tf.name_scope("Layer2"):
W2 = tf.Variable(tf.random_uniform([innerN, 1], minval=-2/512, maxval=2/512), name="Weights_2")
#W2 = tf.Variable(tf.zeros([innerN, 1]), name="Weights_2")
b2 = tf.Variable(tf.zeros([1]), name="Bias_2")
y = tf.nn.sigmoid( tf.matmul(Out1, W2) + b2)
with tf.name_scope("Training"):
y_ = tf.placeholder(tf.float32, [None,1])
cross_entropy = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits(
labels = y_, logits = y)
)
train_step = tf.train.GradientDescentOptimizer(0.005).minimize(cross_entropy)
with tf.name_scope("Testing"):
# Test trained model
correct_prediction = tf.equal( tf.round(y), tf.round(y_))
# ...
# Train
for step in range(500):
batch_xs, batch_ys = Datasets.train.next_batch(300, shuffle=False)
_, my_y, summary = sess.run([train_step, y, merged_summaries],
feed_dict={x: batch_xs, y_: batch_ys})
I suspect two cases:
my fault – bad NN implementation, wrong architecture;
bad data. Compared to XOR example, incomplete training data would result in a failing NN. However, the training examples fed to the trained NN are supposed to give right predictions, aren't they?
How to evaluate if it is possible at all to train a neural network (a 2-layer perceptron) on the provided data to forecast the result? A case of aceptable set would be the XOR example. Opposed to some random noise.
There are only ad hoc ways to know if it is possible to learn a function with a differentiable network from a dataset. That said, these ad hoc ways do usually work. For example, the network should be able to overfit the training set without any regularisation.
A common technique to gauge this is to only fit the network on a subset of the full dataset. Check that the network can overfit to that, then increase the size of the subset, and increase the size of the network as well. Unfortunately, deciding whether to add extra layers or add more units in a hidden layer is an arbitrary decision you'll have to make.
However, looking at your code, there are a few things that could be going wrong here:
Are your outputs balanced? By that I mean, do you have the same number of 1s as 0s in the dataset targets?
Your initialisation in the first layer is all zeros, the gradient to this will be zero, so it can't learn anything (although, you have a real initialisation above it commented out).
Sigmoid nonlinearities are more difficult to optimise than simpler nonlinearities, such as ReLUs.
I'd recommend using the built-in definitions for layers in Tensorflow to not worry about initialisation, and switching to ReLUs in any hidden layers (you need sigmoid at the output for your boolean target).
Finally, deep learning isn't actually very good at most "bag of features" machine learning problems because they lack structure. For example, the order of the features doesn't matter. Other methods often work better, but if you really want to use deep learning then you could look at this recent paper, showing improved performance by just using a very specific nonlinearity and weight initialisation (change 4 lines in your code above).
The output for network which recognizes MNIST database is predictions for 10 classes, i.e. for 1000 images we will have matrix of size (1000, 10). Which way represents cost function better (and why?):
simple tf.squared_difference(Y_pred, Y_pred)?
mean over batches tf.reduce_mean(tf.tf.squared_difference(Y_pred, Y_pred), axis=0)
or mean over classes and then over batches tf.reduce_men(tf.reduce_mean(tf.tf.squared_difference(Y_pred, Y_pred), axis=1))
Thanks
Since your targets are categories (numbers from 0 to 9), it is better to use a cross entropy cost function, and do something like that:
cross_entropy = -tf.reduce_sum(Y * tf.log(Y_pred + 1e-10))
optimizer = tf.train.AdamOptimizer(0.001).minimize(cross_entropy)
Here Y are the real target values, and Y_pred are the predicted target values.
Here is a discussion of the benefits of cross entropy over mean square error for classification problems.