def test_ml_updates():
r = ssrbm.ssRBM(n_h=nh,
n_v=nv,
bw_h=bwh,
iscales={
'Wv': 0.1,
'hbias': 0.,
'alpha': 1.,
'mu': 0.,
'beta': 1.
},
compile=True,
batch_size=batch_size,
sp_weight={'h': 0.},
sp_targ={'h': 0.},
sparse_hmask=sparse_masks.SparsityMask.unfactored_g(
nh, nh, bwh, bwh),
parametrize_sqrt_precision=False,
l1={
'Wh': 0.,
'Wv': 0.
},
l2={
'Wh': 0.,
'Wv': 0.
})
x = numpy.random.rand(batch_size, nv)
r.batch_train_func(x)
def quick_alloc():
r = ssrbm.ssRBM(n_h=nh,
n_v=nv,
bw_h=bwh,
iscales={
'Wv': 0.1,
'hbias': 0.,
'alpha': 1.,
'mu': 0.,
'beta': 1.
},
compile=False,
batch_size=batch_size,
sp_weight={'h': 0.},
sp_targ={'h': 0.},
sparse_hmask=sparse_masks.SparsityMask.unfactored_g(
nh, nh, bwh, bwh),
parametrize_sqrt_precision=False,
l1={
'Wh': 0.,
'Wv': 0.
},
l2={
'Wh': 0.,
'Wv': 0.
})
r.Wv.set_value(Wv)
r.hbias.set_value(hbias)
r.mu.set_value(mu)
r.alpha.set_value(alpha)
r.beta.set_value(beta)
return r
def test_ml_updates():
r = ssrbm.ssRBM(n_h=nh, n_v=nv, bw_h=bwh,
iscales={'Wv': 0.1, 'hbias':0., 'alpha':1., 'mu':0., 'beta':1.},
compile=True, batch_size=batch_size,
sp_weight={'h':0.},
sp_targ={'h':0.},
sparse_hmask = sparse_masks.SparsityMask.unfactored_g(nh,nh,bwh,bwh),
parametrize_sqrt_precision = False,
l1={'Wh':0.,'Wv':0.},
l2={'Wh':0.,'Wv':0.})
x = numpy.random.rand(batch_size, nv)
r.batch_train_func(x)
def quick_alloc():
r = ssrbm.ssRBM(n_h=nh, n_v=nv, bw_h=bwh,
iscales={'Wv': 0.1, 'hbias':0., 'alpha':1., 'mu':0., 'beta':1.},
compile=False, batch_size=batch_size,
sp_weight={'h':0.},
sp_targ={'h':0.},
sparse_hmask = sparse_masks.SparsityMask.unfactored_g(nh,nh,bwh,bwh),
parametrize_sqrt_precision = False,
l1={'Wh':0.,'Wv':0.},
l2={'Wh':0.,'Wv':0.})
r.Wv.set_value(Wv)
r.hbias.set_value(hbias)
r.mu.set_value(mu)
r.alpha.set_value(alpha)
r.beta.set_value(beta)
return r