Skip to content

Fix robust_fit for large timestamps by subtracting offset #73

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
cboulay merged 2 commits into from
Dec 30, 2020
Merged

Fix robust_fit for large timestamps by subtracting offset #73

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
22edf50
Fix robust_fit for large timestamps by subtracting offset
cboulay Dec 29, 2020
710b877
robust_fit: copy mutable input A.
cboulay Dec 29, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 9 additions & 4 deletions pyxdf/pyxdf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -611,11 +611,12 @@ def _clock_sync(
if range_i[0] != range_i[1]:
start, stop = range_i[0], range_i[1] + 1
e = np.ones((stop - start,))
X = np.column_stack([e, clock_times[start:stop]])
X /= winsor_threshold
y = clock_values[start:stop] / winsor_threshold
X = np.column_stack([e, np.array(clock_times[start:stop]) / winsor_threshold])
y = np.array(clock_values[start:stop]) / winsor_threshold
# noinspection PyTypeChecker
coef.append(_robust_fit(X, y))
_coefs = _robust_fit(X, y)
_coefs[0] *= winsor_threshold
coef.append(_coefs)
else:
coef.append((clock_values[range_i[0]], 0))

Expand Down Expand Up @@ -702,6 +703,9 @@ def _robust_fit(A, y, rho=1, iters=1000):
http://www.stanford.edu/~boyd/papers/distr_opt_stat_learning_admm.html

"""
A = np.copy(A) # Don't mutate input.
offset = np.min(A[:, 1])
A[:, 1] -= offset
Aty = np.dot(A.T, y)
L = np.linalg.cholesky(np.dot(A.T, A))
U = L.T
Expand All @@ -715,6 +719,7 @@ def _robust_fit(A, y, rho=1, iters=1000):
tmp = np.maximum(0, (1 - (1 + 1 / rho) * np.abs(d_inv)))
z = rho / (1 + rho) * d + 1 / (1 + rho) * tmp * d
u = d - z
x[0] -= x[1] * offset
return x


Expand Down