def ho_svd(self):
X = square_tensor_gen(self.n,
self.rank,
dim=self.dim,
typ=self.gen_typ,
noise_level=self.noise_level)
start_time = time.time()
core, tucker_factors = tucker(
X, ranks=[self.rank for _ in range(self.dim)], init='random')
X_hat = tl.tucker_to_tensor(core, tucker_factors)
running_time = time.time() - start_time
rerr = eval_mse(X, X_hat)
return (-1, running_time), rerr
def two_pass(self):
X = square_tensor_gen(self.n,
self.rank,
dim=self.dim,
typ=self.gen_typ,
noise_level=self.noise_level)
start_time = time.time()
sketch = Sketch(X, self.k, random_seed=None)
sketchs, _, = sketch.get_sketchs()
sketch_time = time.time() - start_time
start_time = time.time()
sketch_two_pass = SketchTwoPassRecover(X, sketchs, self.rank)
X_hat, _, _ = sketch_two_pass.recover()
recover_time = time.time() - start_time
rerr = eval_mse(X, X_hat)
return (sketch_time, recover_time), rerr
def one_pass(self):
X = square_tensor_gen(self.n,
self.rank,
dim=self.dim,
typ=self.gen_typ,
noise_level=self.noise_level)
start_time = time.time()
sketch = Sketch(X, self.k, s=self.s, random_seed=self.random_seed)
sketchs, core_sketch, = sketch.get_sketchs()
sketch_time = time.time() - start_time
start_time = time.time()
sketch_one_pass = SketchOnePassRecover(sketchs, core_sketch,
self.Tinfo_bucket,
self.Rinfo_bucket)
X_hat, _, _ = sketch_one_pass.recover()
recover_time = time.time() - start_time
rerr = eval_mse(X, X_hat)
return (sketch_time, recover_time), rerr
reduxs2 = [redux1, redux6, redux7]
sims_colspace4 = Simulations(X, reduxs2)
sims_colspace4_result = sims_colspace4.run_colspace_varyk(ks)
pickle.dump(sims_colspace4_result, open("data/col_dim"+str(dim)+"_krao_m"+str(m)+".pickle", "wb" ))
# sims_colspace4_result = pickle.load( open("data/col_dim"+str(dim)+"_krao_m"+str(m)+".pickle", "rb" ) )
# sims_colspace4.plot_colspace_varyk(sims_colspace4_result, ks, ["Gaussian","Gaussian Khatri-Rao", "Gaussian Khatri-Rao Variance Reduced"], str(dim)+"-D "+dim_str(n,dim)+" Tensor Sketching: m = "+str(m),"col_dim"+ str(dim)+"_krao_m"+str(m))
'''
# In[ ]:
k = 30
n = 100
dim = 4
X0 = square_tensor_gen(n, r = 5, dim = dim, typ = 'lk', noise_level=0.1)[0]
X = tl.unfold(X0,mode = 0).T
m = X.shape[0]
krao_ms = np.repeat(int(np.sqrt(n)),2*(dim - 1))
ks = np.arange(5,30,5)
redux1 = DimRedux('g',k, m)
redux2 = DimRedux('u',k, m)
redux3 = DimRedux('sgn',k, m)
redux4 = DimRedux('sp0',k, m)
redux5 = DimRedux('sp1',k, m)
redux6 = DimRedux('g',k, m, krao = True, krao_ms = krao_ms)
redux7 = DimRedux('g',k, m, krao = True, krao_ms = krao_ms, vr = True, vr_typ = 'geom_median_l2')
reduxs1 = [redux1, redux2, redux3, redux4, redux5]
reduxs2 = [redux1, redux6, redux7]
self.core_tensor,
np.linalg.pinv(np.dot(phis[mode_n], Qs[mode_n])),
mode=mode_n)
core_tensor, factors = tucker(self.core_tensor,
ranks=[self.rank for _ in range(dim)])
self.core_tensor = core_tensor
for n in range(dim):
self.arms.append(np.dot(Qs[n], factors[n]))
X_hat = tl.tucker_to_tensor(self.core_tensor, self.arms)
return X_hat, self.arms, self.core_tensor
from util import square_tensor_gen
from sketch import *
if __name__ == "__main__":
tl.set_backend('numpy')
X = square_tensor_gen(10, 3, dim=3, typ='spd', noise_level=0.1)
print(tl.unfold(X, mode=1).shape)
tensor_sketch = Sketch(X,
5,
random_seed=1,
s=-1,
typ='g',
sparse_factor=0.1)
sketchs, core_sketch = tensor_sketch.get_sketchs()
SketchTwoPassRecover(tensor_sketch, np.repeat(3, 3))
# SketchOnePassRecover(tensor_sketch)