FazBrowse GitHub Viewer | Trending |
URL:
| Home
Tools: [Download Repo ZIP]   [Original HTTPS Page]

Fixed #1006 RuntimeWarning `divide by zero encountered in divide` by NimaSarajpoor · Pull Request #1012 · stumpy-dev/stumpy · GitHub

/ stumpy Public

Fixed #1006 RuntimeWarning divide by zero encountered in divide #1012

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

Open
NimaSarajpoor wants to merge 11 commits into
base: main
Choose a base branch
Could not load branches
Branch not found: {{ refName }}

Are you sure you want to change the base?

Some commits from the old base branch may be removed from the timeline, and old review comments may become outdated.
from
Conversation 16 Commits 11 Checks 31 Files changed
Open

Fixed #1006 RuntimeWarning divide by zero encountered in divide #1012

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
47884e6
Add test to catch warning divide-by-zero raised by process_diagonal
NimaSarajpoor Jul 21, 2024
1977825
Add code to resolve warning
NimaSarajpoor Jul 22, 2024
096b332
set std of non-finite subseq to np.nan to avoid divide by 0 warning
NimaSarajpoor Jul 23, 2024
f154e2e
add test for divide-by-zero warning
NimaSarajpoor Jul 25, 2024
f1e250a
add test to capture divide-by-zero warning in ostinato
NimaSarajpoor Jul 25, 2024
09b7fa8
Fix the divide-by-zero warning caused by passing processed data to stump
NimaSarajpoor Jul 25, 2024
4084415
add test for ostinatoed to capture divide by zero warning and fix code
NimaSarajpoor Jul 25, 2024
c5cc377
Rename test functions for sake of consistency
NimaSarajpoor Jul 25, 2024
5e056d3
minor change to fix the computed joint matrix profile
NimaSarajpoor Jul 26, 2024
6cff2f8
merged and resolved conflict
NimaSarajpoor Jul 26, 2026
f7c52d9
minor changes
NimaSarajpoor Jul 27, 2026
File filter

Filter by extension

Filter by extension .py  (5) All 1 file type selected
Viewed files
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Unified
Split
Hide whitespace
Diff view
Unified
Split
Hide whitespace
1 change: 1 addition & 0 deletions stumpy/core.py
Show comments
View file
Open in desktop
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
Show hidden characters
Original file line number Diff line number Diff line change
Expand Up @@ -2286,6 +2286,7 @@ def preprocess_diagonal(

M_T, Σ_T = compute_mean_std(T, m)
Σ_T[T_subseq_isconstant] = 1.0 # Avoid divide by zero in next inversion step
Σ_T[~T_subseq_isfinite] = np.nan
Σ_T_inverse = 1.0 / Σ_T
M_T_m_1, _ = compute_mean_std(T, m - 1)

Expand Down
3 changes: 2 additions & 1 deletion stumpy/gpu_ostinato.py
Show comments
View file
Open in desktop
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
Show hidden characters
Original file line number Diff line number Diff line change
Expand Up @@ -113,9 +113,10 @@ def gpu_ostinato(Ts, m, device_id=0, normalize=True, p=2.0, Ts_subseq_isconstant
M_Ts = [None] * len(Ts)
Σ_Ts = [None] * len(Ts)
for i, T in enumerate(Ts):
Ts_copy[i], M_Ts[i], Σ_Ts[i], Ts_subseq_isconstant[i] = core.preprocess(
_, M_Ts[i], Σ_Ts[i], Ts_subseq_isconstant[i] = core.preprocess(
T, m, copy=True, T_subseq_isconstant=Ts_subseq_isconstant[i]
)
Ts_copy[i] = T.copy()

bsf_radius, bsf_Ts_idx, bsf_subseq_idx = _ostinato(
Ts_copy,
Expand Down
18 changes: 9 additions & 9 deletions stumpy/ostinato.py
Show comments
View file
Open in desktop
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
Show hidden characters
Original file line number Diff line number Diff line change
Expand Up @@ -58,10 +58,9 @@ def _across_series_nearest_neighbors(
nns_subseq_idx = np.zeros(k, dtype=np.int64)

for i in range(k):
QT = core.sliding_dot_product(Ts[Ts_idx][subseq_idx : subseq_idx + m], Ts[i])
distance_profile = core._mass(
Q,
Ts[i],
QT = core.sliding_dot_product(Q, Ts[i])

NimaSarajpoor Jul 26, 2024 •
edited
Loading

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Choose a reason Spam Abuse Off Topic Outdated Duplicate Resolved Low Quality

Since Q and Ts[i] are UNPROCESSED, QT[idx] can get meaningless value for a non-finite subsequence that starts at idx. Better to just replace them with np.nan. So, for example, we can do something like this after computing QT (QT = core.sliding_dot_product(Q, Ts[i])):

# post-processing QT
if np.any(~np.isfinite(Q)):
     QT[:] = np.nan

QT[~np.isfinite(M_Ts[i])] = np.nan

Note that QT[idx] will be ignored though eventually once it comes to computing its corresponding value in the distance profile via core.calculate_distance_profile (see a few lines below). In other words, we do not need the "post-processing of QT" but I think it makes code clearer.

Any concerns?

[Update]
Note: Since we pass inputs with non-finite values to core.calculate_distance_profile (Not only QT but also M_T), maybe I should better stop here and try to address #1011 first. Then come back and resume the work on this PR.

seanlaw Jul 26, 2024

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Choose a reason Spam Abuse Off Topic Outdated Duplicate Resolved Low Quality

I understand the need to replace the ugly Ts[Ts_idx][subseq_idx : subseq_idx + m] with Q. That's fine but

QT = core.sliding_dot_product(Q, Ts[i])

doesn't seem to require any post-processing when either Q or T contain np.nan.

NimaSarajpoor Jul 27, 2024 •
edited
Loading

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Choose a reason Spam Abuse Off Topic Outdated Duplicate Resolved Low Quality

My bad! You are right. No need for post-processing QT

NimaSarajpoor Aug 3, 2024

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Choose a reason Spam Abuse Off Topic Outdated Duplicate Resolved Low Quality

@seanlaw

maybe I should better stop here and try to address #1011 first. Then come back and resume the work on this PR.

I did an experiment and shared the result in #1011. We can either work on that and then come back here. Or, on the second thought, we can still merge this (after I resolve the merge conflicts), and then work on #1011.

What do you think?

seanlaw Aug 3, 2024

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Choose a reason Spam Abuse Off Topic Outdated Duplicate Resolved Low Quality

Since we have a merge conflict here anyways, let's fix #1011 first and then come back here and strikethrough the comment:

Note: Since we pass inputs with non-finite values to core.calculate_distance_profile (Not only QT but also M_T), maybe I should better stop here and try to address #1011 first. Then come back and resume the work on this PR.

so that we are focused on this issue/PR instead of being distracted by jumping back and forth. I am getting confused by having so many things open and so it feels disorganized
distance_profile = core.calculate_distance_profile(
m,
QT,
M_Ts[Ts_idx][subseq_idx],
Σ_Ts[Ts_idx][subseq_idx],
Expand Down Expand Up @@ -256,9 +255,8 @@ def _ostinato(
(
radius,
np.min(
core._mass(
Ts[j][q : q + m],
Ts[i],
core.calculate_distance_profile(
m,
QT,
M_Ts[j][q],
Σ_Ts[j][q],
Expand Down Expand Up @@ -373,9 +371,10 @@ def ostinato(Ts, m, normalize=True, p=2.0, Ts_subseq_isconstant=None):
Ts_subseq_isconstant = [None] * len(Ts)

for i, T in enumerate(Ts):
Ts_copy[i], M_Ts[i], Σ_Ts[i], Ts_subseq_isconstant[i] = core.preprocess(
_, M_Ts[i], Σ_Ts[i], Ts_subseq_isconstant[i] = core.preprocess(
T, m, copy=True, T_subseq_isconstant=Ts_subseq_isconstant[i]
)
Ts_copy[i] = T.copy()

bsf_radius, bsf_Ts_idx, bsf_subseq_idx = _ostinato(
Ts_copy, m, M_Ts, Σ_Ts, Ts_subseq_isconstant
Expand Down Expand Up @@ -516,9 +515,10 @@ def ostinatoed(client, Ts, m, normalize=True, p=2.0, Ts_subseq_isconstant=None):
if Ts_subseq_isconstant is None:
Ts_subseq_isconstant = [None] * len(Ts)
for i, T in enumerate(Ts):
Ts_copy[i], M_Ts[i], Σ_Ts[i], Ts_subseq_isconstant[i] = core.preprocess(
_, M_Ts[i], Σ_Ts[i], Ts_subseq_isconstant[i] = core.preprocess(
T, m, copy=True, T_subseq_isconstant=Ts_subseq_isconstant[i]
)
Ts_copy[i] = T.copy()

bsf_radius, bsf_Ts_idx, bsf_subseq_idx = _ostinato(
Ts_copy,
Expand Down
12 changes: 12 additions & 0 deletions tests/test_core.py
Show comments
View file
Open in desktop
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
Show hidden characters
Original file line number Diff line number Diff line change
Expand Up @@ -881,9 +881,11 @@ def test_preprocess_non_normalized():
def test_preprocess_diagonal():
T = np.array([0, np.nan, 2, 3, 4, 5, 6, 7, np.inf, 9])
m = 3
T_subseq_isfinite = core.rolling_isfinite(T, m)

ref_T = np.array([0, 0, 2, 3, 4, 5, 6, 7, 0, 9], dtype=float)
ref_M, ref_Σ = naive.compute_mean_std(ref_T, m)
ref_Σ[~T_subseq_isfinite] = np.nan
ref_Σ_inverse = 1.0 / ref_Σ
ref_M_m_1, _ = naive.compute_mean_std(ref_T, m - 1)

Expand Down Expand Up @@ -1758,6 +1760,16 @@ def test_process_isconstant_2d():
npt.assert_array_equal(T_subseq_isconstant_ref, T_subseq_isconstant_comp)


@pytest.mark.filterwarnings(
"error:divide by zero encountered in divide", category=RuntimeWarning
)
def test_preprocess_diagonal_divide_by_zero():
T = np.random.rand(64)
m = 3
T[:m] = np.nan
core.preprocess_diagonal(T, m)


def test_process_isconstant_1d_default():
# test the default value of `T_subseq_isconstant` in `process_isconstant`
n = 64
Expand Down
21 changes: 21 additions & 0 deletions tests/test_ostinato.py
Show comments
View file
Open in desktop
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
Show hidden characters
Original file line number Diff line number Diff line change
Expand Up @@ -248,3 +248,24 @@ def test_extract_several_consensus_ostinatoed(dask_cluster):
Ts_ref[i][np.isfinite(Ts_ref[i])],
Ts_comp[i][np.isfinite(Ts_comp[i])],
)


@pytest.mark.filterwarnings(
"error:divide by zero encountered in divide", category=RuntimeWarning
)
def test_divide_by_zero_ostinato():
Ts = [np.random.rand(n) for n in [64, 128, 256]]
m = 5
Ts[0][:m] = np.nan
ostinato(Ts, m)


@pytest.mark.filterwarnings(
"error:divide by zero encountered in divide", category=RuntimeWarning
)
def test_divide_by_zero_ostinatoed(dask_cluster):
Ts = [np.random.rand(n) for n in [64, 128, 256]]
m = 5
Ts[0][:m] = np.nan
with Client(dask_cluster) as dask_client:
ostinatoed(dask_client, Ts, m)
Loading
Back | FazBrowse Home | New Git URL