quaternion_regression.py

import numpy
import pandas
from keras.models import Sequential
from keras.layers import Dense
from keras import optimizers
from keras import regularizers

# load dataset
dataframe = pandas.read_csv("some_quaternions.csv", delim_whitespace=False, header=None)
dataset = dataframe.values
# expected == input
X = dataset[:-10]
Y = dataset[:-10]

model = Sequential()
model.add(Dense(20, input_dim=4, 
					kernel_initializer='normal',
					kernel_regularizer=regularizers.l2(0.01), 
					activation='relu'))

model.add(Dense(4, kernel_initializer='normal'))

opt = optimizers.Adam(lr=0.001, beta_1=0.9, beta_2=0.999, decay=1e-6)
model.compile(loss='mean_squared_error', optimizer=opt)

model.fit(X, Y, epochs=50, batch_size=10)

##########

predictions = model.predict(dataset[-10:])

for i,row in enumerate(predictions):
	print "PRED {}:".format(i)

	print row
	print dataset[-10:].reshape(10,4)[i]

some_quaternions.csv