import numpy as np
import causet as cs
#testing that the causalmatrix is correctly calculated by causDiam
#this is not the most efficient way to calculate the entries of c and tests if
#the vectorization in its implementation in causDiam is done correctly
d = 4
c, t, x = cs.causDiam(1000, d)
ds2 = 0.
finalTruth = True
for i in range(1000):
for j in range(1000):
ds2 = -(t[0, j] - t[0, i])**2
ds2 = ds2 + (x[0, j] - x[0, i])**2
if (d == 3) or (d == 4):
ds2 = ds2 + (x[1, j] - x[1, i])**2
if d == 4:
ds2 = ds2 + (x[2, j] - x[2, i])**2
tempBool = (ds2 < 0) and (t[0, i] < t[0, j])
finalTruth = (finalTruth and (tempBool == c[i, j]))
#finalTruth should be True at the end of these loops
# generating the data to be used in the training of the autoencoder
# for now we are generating causets in flat space in d dimensions
# later we may incorporate other features such as curvature
tot_train = 2000 #total number of causal sets (in training set) generated per dimension
tot_test = 500 #total number of causal sets (in test set) generated per dimension
n = 100 #total sprinkled points in causet
dims = np.array([2, 3, 4])
xtrain = np.zeros((tot_train * dims.size, n**2), dtype='int8')
ytrain = np.zeros((tot_train * dims.size, ), dtype='int8')
xtest = np.zeros((tot_test * dims.size, n**2), dtype='int8')
ytest = np.zeros((tot_test * dims.size, ), dtype='int8')
for d in range(dims.size):
for i in range(tot_train):
c, t, x = cs.causDiam(n, dims[d])
xtrain[d * tot_train + i] = c.reshape(1, n**2)
ytrain[d * tot_train + i] = dims[d]
for i in range(tot_test):
c, t, x = cs.causDiam(n, dims[d])
xtest[d * tot_test + i] = c.reshape(1, n**2)
ytest[d * tot_test + i] = dims[d]
np.save('xtrain.npy', xtrain)
np.save('ytrain.npy', ytrain)
np.save('xtest.npy', xtest)
np.save('ytest.npy', ytest)