def _flatten_structures(self):
c_specs, self.edges, self.weights, self.einstrs, self.einpaths = [], [], [], [], []
ks = {}
# handle n=1 case specially
c_specs.append([1,0,0,0,0,1,1,0])
self.edges.append(())
self.weights.append(())
self.einstrs.append(self.einstrs_d[(1,0)][0])
self.einpaths.append(self.einpaths_d[(1,0)][0])
for ne in sorted(self.edges_d.keys()):
n, e = ne
z = zip(self.edges_d[ne], self.weights_d[ne], self.chis_d[ne],
self.einstrs_d[ne], self.einpaths_d[ne])
for edgs, ws, c, es, ep in z:
for w in ws:
d = sum(w)
k = ks.setdefault((n,d), 0)
ks[(n,d)] += 1
vs = valencies(EFPElem(edgs, weights=w).edges).values()
v = max(vs)
h = Counter(vs)[1]
c_specs.append([n, e, d, v, k, c, 1, h])
self.edges.append(edgs)
self.weights.append(w)
self.einstrs.append(es)
self.einpaths.append(ep)
self.c_specs = np.asarray(c_specs)
def __init__(self,
edges,
measure='hadr',
beta=1,
kappa=1,
normed=True,
coords=None,
check_input=True,
np_optimize='greedy'):
"""
**Arguments**
- **edges** : _list_
- Edges of the EFP graph specified by pairs of vertices.
- **measure** : {`'hadr'`, `'hadrdot'`, `'ee'`}
- The choice of measure. See [Measures](../measures)
for additional info.
- **beta** : _float_
- The parameter $\\beta$ appearing in the measure.
Must be greater than zero.
- **kappa** : {_float_, `'pf'`}
- If a number, the energy weighting parameter $\\kappa$.
If `'pf'`, use $\\kappa=v-1$ where $v$ is the valency of the vertex.
- **normed** : _bool_
- Controls normalization of the energies in the measure.
- **coords** : {`'ptyphim'`, `'epxpypz'`, `None`}
- Controls which coordinates are assumed for the input. See
[Measures](../measures) for additional info.
- **check_input** : _bool_
- Whether to check the type of the input each time or assume
the first input type.
- **np_optimize** : {`True`, `False`, `'greedy'`, `'optimal'`}
- The `optimize` keyword of `numpy.einsum_path`.
"""
# initialize EFPBase
super(EFP, self).__init__(measure, beta, kappa, normed, coords,
check_input)
# store these edges as an EFPElem
self.efpelem = EFPElem(edges)
self._np_optimize = np_optimize
# setup ve for standard efp compute
(self.efpelem.einstr, self.efpelem.einpath,
self.chi) = VariableElimination(np_optimize).einspecs(
self.simple_graph, self.n)
def __init__(self, *args, **kwargs):
r"""EFPSet can be initialized in one of three ways (in order of precedence):
1. **Default** - Use the ($d\le10$) EFPs that come installed with the
`EnergFlow` package.
2. **Generator** - Pass in a custom `Generator` object as the
first positional argument.
3. **Custom File** - Pass in the name of a `.npz` file saved
with a custom `Generator`.
To control which EFPs are included, `EFPSet` accepts an arbitrary
number of specifications (see [`sel`](#sel)) and only EFPs meeting each
specification are included in the set.
**Arguments**
- ***args** : _arbitrary positional arguments_
- If the first positional argument is a `Generator` instance,
it is used for initialization. The remaining positional
arguments must be valid arguments to `sel`.
- **filename** : _string_
- Path to a `.npz` file which has been saved by a valid
`energyflow.Generator`.
- **measure** : {`'hadr'`, `'hadr-dot'`, `'ee'`}
- See [Measures](../measures) for additional info.
- **beta** : _float_
- The parameter $\\beta$ appearing in the measure.
Must be greater than zero.
- **kappa** : {_float_, `'pf'`}
- If a number, the energy weighting parameter $\\kappa$.
If `'pf'`, use $\\kappa=v-1$ where $v$ is the valency of the vertex.
- **normed** : _bool_
- Controls normalization of the energies in the measure.
- **coords** : {`'ptyphim'`, `'epxpypz'`, `None`}
- Controls which coordinates are assumed for the input. See
[Measures](../measures) for additional info.
- **check_input** : _bool_
- Whether to check the type of the input each time or assume
the first input type.
- **verbose** : _bool_
- Controls printed output when initializing EFPSet.
"""
default_kwargs = {
'filename': None,
'measure': 'hadr',
'beta': 1,
'kappa': 1,
'normed': True,
'coords': None,
'check_input': True,
'verbose': False
}
measure_kwargs = [
'measure', 'beta', 'kappa', 'normed', 'coords', 'check_input'
]
# process arguments
for k, v in default_kwargs.items():
if k not in kwargs:
kwargs[k] = v
if k not in measure_kwargs:
setattr(self, k, kwargs.pop(k))
kwargs_check('__init__', kwargs, allowed=measure_kwargs)
# initialize EFPBase
super(EFPSet, self).__init__(*[kwargs[k] for k in measure_kwargs])
# handle different methods of initialization
maxs = ['nmax', 'emax', 'dmax', 'cmax', 'vmax', 'comp_dmaxs']
elemvs = ['edges', 'weights', 'einstrs', 'einpaths']
if len(args) >= 1 and isinstance(args[0], Generator):
constructor_attrs = maxs + elemvs + [
'cols', 'c_specs', 'disc_specs', 'disc_formulae'
]
gen = {attr: getattr(args[0], attr) for attr in constructor_attrs}
args = args[1:]
elif self.filename is not None:
self.filename += '.npz' if not self.filename.endswith(
'.npz') else ''
gen = np.load(self.filename, allow_pickle=True)
else:
gen = np.load(DEFAULT_EFP_FILE, allow_pickle=True)
# compile regular expression for use in sel()
self.SEL_RE = SEL_RE
# put column headers and indices into namespace
self._cols = gen['cols']
self._set_col_inds()
# put gen maxs into dict
self.gen_maxs = {m: gen[m] for m in maxs}
# get disc formulae and disc mask
orig_disc_specs = gen['disc_specs']
disc_mask = self.sel(*args, specs=orig_disc_specs)
self.disc_formulae = gen['disc_formulae'][disc_mask]
# get connected specs and full specs
orig_c_specs = gen['c_specs']
c_mask = self.sel(*args, specs=orig_c_specs)
self._cspecs = orig_c_specs[c_mask]
self._specs = concat_specs(self._cspecs, orig_disc_specs[disc_mask])
# make EFPElem list
z = zip(*([gen[v] for v in elemvs] + [orig_c_specs[:, self.k_ind]]))
self.efpelems = [
EFPElem(*args) for m, args in enumerate(z) if c_mask[m]
]
# union over all weights needed
self.__weight_set = frozenset(w for efpelem in self.efpelems
for w in efpelem.weight_set)
# get col indices for disconnected formulae
connected_ndk = {
efpelem.ndk: i
for i, efpelem in enumerate(self.efpelems)
}
self.disc_col_inds = []
for formula in self.disc_formulae:
try:
self.disc_col_inds.append(
[connected_ndk[factor] for factor in formula])
except KeyError:
warnings.warn(
'connected efp needed for {} not found'.format(formula))
# handle printing
if self.verbose:
print('Originally Available EFPs:')
self.print_stats(specs=concat_specs(orig_c_specs, orig_disc_specs),
lws=2)
if len(args) > 0:
print('Currently Stored EFPs:')
self.print_stats(lws=2)