lattice/pyerrors
1
0
Fork 0
mirror of https://github.com/fjosw/pyerrors.git synced 2025-05-14 19:43:41 +02:00
Code Issues Projects Releases Packages Wiki Activity

Test version to investigate recentering the deltas

Browse source
This commit is contained in:
Simon Kuberski 2023-02-02 11:21:26 +01:00
parent cb31e681b8
commit 74fb9a0a74
1 changed files with 15 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

18
pyerrors/obs.py
View file

@ -1073,7 +1073,7 @@ def _intersection_idx(idl):
return sorted(set.intersection(*[set(o) for o in idl])) return sorted(set.intersection(*[set(o) for o in idl]))
def _expand_deltas_for_merge(deltas, idx, shape, new_idx): def _expand_deltas_for_merge(deltas, idx, shape, new_idx, shift):
"""Expand deltas defined on idx to the list of configs that is defined by new_idx. """Expand deltas defined on idx to the list of configs that is defined by new_idx.
New, empty entries are filled by 0. If idx and new_idx are of type range, the smallest New, empty entries are filled by 0. If idx and new_idx are of type range, the smallest
common divisor of the step sizes is used as new step size. common divisor of the step sizes is used as new step size.
@ -1089,12 +1089,16 @@ def _expand_deltas_for_merge(deltas, idx, shape, new_idx):
Number of configs in idx. Number of configs in idx.
new_idx : list new_idx : list
List of configs that defines the new range, has to be sorted in ascending order. List of configs that defines the new range, has to be sorted in ascending order.
shift: double
Shift that is applied to each delta to properly recenter around the mean.
""" """
if type(idx) is range and type(new_idx) is range: if type(idx) is range and type(new_idx) is range:
if idx == new_idx: if idx == new_idx:
return deltas return deltas
ret = np.zeros(new_idx[-1] - new_idx[0] + 1) ret = np.zeros(new_idx[-1] - new_idx[0] + 1)
if abs(shift) > np.finfo(np.float64).eps:
deltas = np.array(deltas) + shift
for i in range(shape): for i in range(shape):
ret[idx[i] - new_idx[0]] = deltas[i] ret[idx[i] - new_idx[0]] = deltas[i]
return np.array([ret[new_idx[i] - new_idx[0]] for i in range(len(new_idx))]) * len(new_idx) / len(idx) return np.array([ret[new_idx[i] - new_idx[0]] for i in range(len(new_idx))]) * len(new_idx) / len(idx)
@ -1165,6 +1169,7 @@ def derived_observable(func, data, array_mode=False, **kwargs):
new_r_values = {} new_r_values = {}
new_idl_d = {} new_idl_d = {}
shift_d = {}
for name in new_sample_names: for name in new_sample_names:
idl = [] idl = []
tmp_values = np.zeros(n_obs) tmp_values = np.zeros(n_obs)
@ -1175,6 +1180,9 @@ def derived_observable(func, data, array_mode=False, **kwargs):
idl.append(tmp_idl) idl.append(tmp_idl)
if multi > 0: if multi > 0:
tmp_values = np.array(tmp_values).reshape(data.shape) tmp_values = np.array(tmp_values).reshape(data.shape)
shiftav = sum([item.r_values.get(name, item.value) * item.shape.get(name, 0) for item in raveled_data]) / sum([item.shape.get(name, 0) for item in raveled_data])
print(name, np.array([item.shape.get(name, 0) for item in raveled_data]) / sum([item.shape.get(name, 0) for item in raveled_data]))
shift_d[name] = [item.r_values[name] - shiftav if name in item.r_values else 0 for item in raveled_data]
new_r_values[name] = func(tmp_values, **kwargs) new_r_values[name] = func(tmp_values, **kwargs)
new_idl_d[name] = _merge_idx(idl) new_idl_d[name] = _merge_idx(idl)
@ -1215,7 +1223,7 @@ def derived_observable(func, data, array_mode=False, **kwargs):
d_extracted[name] = [] d_extracted[name] = []
ens_length = len(new_idl_d[name]) ens_length = len(new_idl_d[name])
for i_dat, dat in enumerate(data): for i_dat, dat in enumerate(data):
d_extracted[name].append(np.array([_expand_deltas_for_merge(o.deltas.get(name, np.zeros(ens_length)), o.idl.get(name, new_idl_d[name]), o.shape.get(name, ens_length), new_idl_d[name]) for o in dat.reshape(np.prod(dat.shape))]).reshape(dat.shape + (ens_length, ))) d_extracted[name].append(np.array([_expand_deltas_for_merge(o.deltas.get(name, np.zeros(ens_length)), o.idl.get(name, new_idl_d[name]), o.shape.get(name, ens_length), new_idl_d[name]) for o in dat.reshape(np.prod(dat.shape))], 0).reshape(dat.shape + (ens_length, )))
for name in new_cov_names: for name in new_cov_names:
g_extracted[name] = [] g_extracted[name] = []
zero_grad = _Zero_grad(new_covobs_lengths[name]) zero_grad = _Zero_grad(new_covobs_lengths[name])
@ -1241,7 +1249,11 @@ def derived_observable(func, data, array_mode=False, **kwargs):
if name in obs.cov_names: if name in obs.cov_names:
new_grad[name] = new_grad.get(name, 0) + deriv[i_val + j_obs] * obs.covobs[name].grad new_grad[name] = new_grad.get(name, 0) + deriv[i_val + j_obs] * obs.covobs[name].grad
else: else:
new_deltas[name] = new_deltas.get(name, 0) + deriv[i_val + j_obs] * _expand_deltas_for_merge(obs.deltas[name], obs.idl[name], obs.shape[name], new_idl_d[name]) shift = shift_d[name][j_obs[0]]
#shift = deriv[i_val + j_obs] * obs.r_values[name] - new_values[i_val]
print(shift_d, j_obs, shift_d[name][j_obs[0]],)
new_deltas[name] = new_deltas.get(name, 0) + deriv[i_val + j_obs] * _expand_deltas_for_merge(obs.deltas[name], obs.idl[name], obs.shape[name], new_idl_d[name], shift)
#new_deltas[name] = new_deltas.get(name, 0) + deriv[i_val + j_obs] * _expand_deltas_for_merge(obs.deltas[name], obs.idl[name], obs.shape[name], new_idl_d[name], 0)
new_covobs = {name: Covobs(0, allcov[name], name, grad=new_grad[name]) for name in new_grad} new_covobs = {name: Covobs(0, allcov[name], name, grad=new_grad[name]) for name in new_grad}