def test_generate_data(self):
if self.comm.Get_rank() == 0:
self.mu = [np.repeat(np.double(k), self.d) for k in range(self.K)]
r = np.zeros(self.d)
r[0] = 3.
r[1] = -1
self.sigma = [spl.toeplitz(r) for k in range(self.K)]
self.data_sigma = [[
invwishartrand(5, self.sigma[k]) for k in range(self.K)
] for n in range(self.n)] # @UnusedVariable
self.data_mu = [[
self.mu[k] + np.random.randn(self.d) * 0.2
for k in range(self.K)
] for n in range(self.n)] # @UnusedVariable
self.Y = []
for i in range(self.n):
mix = mixture(K=self.K)
mix.mu = self.data_mu[i]
mix.sigma = self.data_sigma[i]
mix.p = np.ones(self.K) / self.K
mix.d = self.d
self.Y.append(mix.simulate_data(self.n_obs))
else:
self.Y = None
self.hGMM.set_data(self.Y)
self.hGMM.set_prior_param0()
def setdata(self): self.mu = [np.repeat(np.double(k),self.d) for k in range(self.K)] self.r1 = np.zeros(self.d) self.r1[0] = 3. self.r1[1] = -1 self.r1[2] = 1 self.r1[2] = .5 self.r2 = np.zeros(self.d) self.r2[0] = 2. self.r2[1] = -1.5 self.r2[2] = 0 self.r2[2] = 0. self.sigma =[np.eye(self.d), spl.toeplitz(self.r1), spl.toeplitz(self.r1), 0.5*np.eye(self.d)] self.p = np.array([1.,1.,1.,0.1]) self.p2 = np.array([1.,1.,1.]) self.p /= np.sum(self.p) self.p2 /= np.sum(self.p2) self.data_sigma = [ [invwishartrand(5, self.sigma[k]) for k in range(self.K)] for n in range(self.n_y)] # @UnusedVariable self.data_mu = [ [self.mu[k]+ np.random.randn(self.d)*0.2 for k in range(self.K)] for n in range(self.n_y)] # @UnusedVariable self.Y = [] for i in range(self.n_y): if i < 17: mix = mixture(K = self.K) mix.mu = self.data_mu[i] mix.sigma = self.data_sigma[i] mix.p = self.p else: mix = mixture(K = self.K-1) mix.mu = self.data_mu[i][:-1] mix.sigma = self.data_sigma[i][:-1] mix.p = self.p2 mix.d = self.d self.Y.append(mix.simulate_data(self.n_obs)) self.hGMM.set_data(self.Y) self.hGMM.set_prior_param0() self.hGMM.set_p_labelswitch(0.5) self.hGMM.set_p_activation([0,0])
def setdata(self): if MPI.COMM_WORLD.Get_rank() == 0: # @UndefinedVariable self.mu = [np.repeat(np.double(k),self.d) for k in range(self.K)] self.r1 = np.zeros(self.d) self.r1[0] = 3. self.r1[1] = -1 self.r1[2] = 1 self.r1[2] = .5 self.r2 = np.zeros(self.d) self.r2[0] = 2. self.r2[1] = -1.5 self.r2[2] = 0 self.r2[2] = 0. self.sigma =[np.eye(self.d), spl.toeplitz(self.r1), spl.toeplitz(self.r1), 0.5*np.eye(self.d)] self.p = np.array([1.,1.,1.,0.1]) self.p2 = np.array([1.,1.,1.]) self.p /= np.sum(self.p) self.p2 /= np.sum(self.p2) self.data_sigma = [ [invwishartrand(5, self.sigma[k]) for k in range(self.K)] for n in range(self.n_y)] # @UnusedVariable self.data_mu = [ [self.mu[k]+ np.random.randn(self.d)*0.2 for k in range(self.K)] for n in range(self.n_y)] # @UnusedVariable self.Y = [] for i in range(self.n_y): if i < 17: mix = mixture(K = self.K) mix.mu = self.data_mu[i] mix.sigma = self.data_sigma[i] mix.p = self.p else: mix = mixture(K = self.K-1) mix.mu = self.data_mu[i][:-1] mix.sigma = self.data_sigma[i][:-1] mix.p = self.p2 mix.d = self.d self.Y.append(mix.simulate_data(self.n_obs)) else: self.Y = None
def generate_data(n_obs, K):
d = 4
n_y = 20
mu = [np.repeat(np.double(k), d) for k in range(K)]
r1 = np.zeros(d)
r1[0] = 3.
r1[1] = -1
r1[2] = 1
r1[2] = .5
r2 = np.zeros(d)
r2[0] = 2.
r2[1] = -1.5
r2[2] = 0
r2[2] = 0.
sigma = [np.eye(d), spl.toeplitz(r1), spl.toeplitz(r1), 0.5 * np.eye(d)]
p = np.array([1., 1., 1., 0.1])
p2 = np.array([1., 1., 1.])
p /= np.sum(p)
p2 /= np.sum(p2)
data_sigma = [[invwishartrand(10 + d, sigma[k]) for k in range(K)]
for n in range(n_y)] # @UnusedVariable
data_mu = [[mu[k] + np.random.randn(d) * 0.005 for k in range(K)]
for n in range(n_y)] # @UnusedVariable
Y = []
for i in range(n_y):
if i < 17:
mix = mixture(K=K)
mix.mu = data_mu[i]
mix.sigma = data_sigma[i]
mix.p = p
else:
mix = mixture(K=K - 1)
mix.mu = data_mu[i][:-1]
mix.sigma = data_sigma[i][:-1]
mix.p = p2
mix.d = d
Y.append(mix.simulate_data(n_obs))
return Y, data_sigma, data_mu
def generate_data( n_obs, K): d = 4 n_y = 20 mu = [np.repeat(np.double(k),d) for k in range(K)] r1 = np.zeros(d) r1[0] = 3. r1[1] = -1 r1[2] = 1 r1[2] = .5 r2 = np.zeros(d) r2[0] = 2. r2[1] = -1.5 r2[2] = 0 r2[2] = 0. sigma =[np.eye(d), spl.toeplitz(r1), spl.toeplitz(r1), 0.5*np.eye(d)] p = np.array([1.,1.,1.,0.1]) p2 = np.array([1.,1.,1.]) p /= np.sum(p) p2 /= np.sum(p2) data_sigma = [ [invwishartrand(10+d, sigma[k]) for k in range(K)] for n in range(n_y)] # @UnusedVariable data_mu = [ [mu[k]+ np.random.randn(d)*0.005 for k in range(K)] for n in range(n_y)] # @UnusedVariable Y = [] for i in range(n_y): if i < 17: mix = mixture(K = K) mix.mu = data_mu[i] mix.sigma = data_sigma[i] mix.p = p else: mix = mixture(K = K-1) mix.mu = data_mu[i][:-1] mix.sigma = data_sigma[i][:-1] mix.p = p2 mix.d = d Y.append(mix.simulate_data(n_obs)) return Y, data_sigma, data_mu
def generate_data(self, p ): self.mu = [np.repeat(np.double(k),self.d) for k in range(self.K)] self.sigma =[np.eye(self.d)*0.01 for k in range (self.K)] mix = mixture(K = self.K) mix.mu = self.mu mix.sigma = self.sigma mix.p = p mix.d = self.d self.Y = mix.simulate_data(self.n_obs) self.mix.mu = self.mu self.mix.sigma = self.sigma self.mix.p = p self.mix.d = self.d self.mix.set_data(self.Y)
def generate_data_mu(self ): p = np.ones(self.K)/self.K self.mu = [np.repeat(np.double(k),self.d) for k in range(self.K)] self.sigma =[np.eye(self.d)*0.01 for k in range (self.K)] mix = mixture(K = self.K) mix.mu = self.mu mix.sigma = self.sigma mix.p = p mix.d = self.d self.Y = mix.simulate_data(self.n_obs) self.mix.mu = cp.deepcopy(self.mu) self.mix.sigma = cp.deepcopy(self.sigma) self.mix.p =cp.deepcopy( p) self.mix.d = self.d self.mix.set_data(self.Y)
def generate_data_mu(self):
p = np.ones(self.K) / self.K
self.mu = [np.repeat(np.double(k), self.d) for k in range(self.K)]
self.sigma = [np.eye(self.d) * 0.01 for k in range(self.K)]
mix = mixture(K=self.K)
mix.mu = self.mu
mix.sigma = self.sigma
mix.p = p
mix.d = self.d
self.Y = mix.simulate_data(self.n_obs)
self.mix.mu = cp.deepcopy(self.mu)
self.mix.sigma = cp.deepcopy(self.sigma)
self.mix.p = cp.deepcopy(p)
self.mix.d = self.d
self.mix.set_data(self.Y)
def generate_data(self, p):
self.mu = [np.repeat(np.double(k), self.d) for k in range(self.K)]
self.sigma = [np.eye(self.d) * 0.01 for k in range(self.K)]
mix = mixture(K=self.K)
mix.mu = self.mu
mix.sigma = self.sigma
mix.p = p
mix.d = self.d
self.Y = mix.simulate_data(self.n_obs)
self.mix.mu = self.mu
self.mix.sigma = self.sigma
self.mix.p = p
self.mix.d = self.d
self.mix.set_data(self.Y)
def generate_data_sigma(self ): p = np.ones(self.K)/self.K self.mu = [np.repeat(np.double(k),self.d) for k in range(self.K-1)] self.mu.append(np.repeat(np.double(0),self.d) ) self.sigma =[spl.toeplitz(self.r)*0.01 for k in range (self.K-1)] self.sigma.append(np.eye(self.d)*0.01 ) mix = mixture(K = self.K) mix.mu = self.mu mix.sigma = self.sigma mix.p = p mix.d = self.d self.Y = mix.simulate_data(self.n_obs) self.mix.mu = cp.deepcopy(self.mu) self.mix.sigma = cp.deepcopy(self.sigma) self.mix.p =cp.deepcopy( p) self.mix.d = self.d self.mix.set_data(self.Y)
def generate_data_sigma(self):
p = np.ones(self.K) / self.K
self.mu = [np.repeat(np.double(k), self.d) for k in range(self.K - 1)]
self.mu.append(np.repeat(np.double(0), self.d))
self.sigma = [spl.toeplitz(self.r) * 0.01 for k in range(self.K - 1)]
self.sigma.append(np.eye(self.d) * 0.01)
mix = mixture(K=self.K)
mix.mu = self.mu
mix.sigma = self.sigma
mix.p = p
mix.d = self.d
self.Y = mix.simulate_data(self.n_obs)
self.mix.mu = cp.deepcopy(self.mu)
self.mix.sigma = cp.deepcopy(self.sigma)
self.mix.p = cp.deepcopy(p)
self.mix.d = self.d
self.mix.set_data(self.Y)
def test_generate_data(self): if self.comm.Get_rank() == 0: self.mu = [np.repeat(np.double(k),self.d) for k in range(self.K)] r = np.zeros(self.d) r[0] = 3. r[1] = -1 self.sigma = [spl.toeplitz(r) for k in range(self.K)] self.data_sigma = [ [invwishartrand(5, self.sigma[k]) for k in range(self.K)] for n in range(self.n)] # @UnusedVariable self.data_mu = [ [self.mu[k]+ np.random.randn(self.d)*0.2 for k in range(self.K)] for n in range(self.n)] # @UnusedVariable self.Y = [] for i in range(self.n): mix = mixture(K = self.K) mix.mu = self.data_mu[i] mix.sigma = self.data_sigma[i] mix.p = np.ones(self.K) / self.K mix.d = self.d self.Y.append(mix.simulate_data(self.n_obs)) else: self.Y = None self.hGMM.set_data(self.Y) self.hGMM.set_prior_param0()