• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

yzhao062 / pyod / 30678244913
93%
master: 93%

Build:
Build:
LAST BUILD BRANCH: add-forte
DEFAULT BRANCH: master
Ran 01 Aug 2026 01:51AM UTC
Jobs 11
Files 189
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

01 Aug 2026 01:34AM UTC coverage: 92.423% (+0.03%) from 92.397%
30678244913

push

github

yzhao062
Fix DeepSVDD hypersphere collapse and center lifecycle (builds on #704)

#704 restored the commented-out loss.backward(), which exposed three
pre-existing defects it had been masking:

- fit() set self.c = 0.0 while _init_c() wrote the computed center to the
  inner module, so the objective never saw it. For a bias-free ReLU network
  the all-zero weights map every input to 0, making c=0 exactly the trivial
  solution of Ruff et al., ICML 2018, Proposition 1. On a 17-dataset ODDS
  benchmark the merged state returned a single distinct score on 10 of 17
  datasets while still clearing a ROC floor via 1e-24 floating-point noise.
- The optimizer was built with no learning rate (Adam default 1e-3) and
  passed l2_regularizer=0.1 as weight_decay, five orders of magnitude above
  the reference 5e-7; that decay drives weights toward zero and compounds
  the collapse.
- best_model_dict was stored but never loaded back, so scoring used
  final-epoch weights.

fit() also wrote the fitted center into the constructor parameter c, which
leaked fitted state through get_params(), made clone() start pre-seeded, and
made a refit train a newly built network against the previous fit's center.

Changes: the fitted center now lives in c_ while c stays configuration; a
user-supplied center is validated (width, finiteness, not all zeros) and
copied so a later mutation cannot change the fitted estimator; l2_regularizer
defaults to 5e-7 and a new learning_rate defaults to 1e-4 (both match the
reference implementation; the l2_regularizer default change is API-visible);
the best-epoch snapshot is applied before scoring.

Measured on the 17-dataset benchmark: mean ROC AUC 0.601 -> 0.748 with score
collapse eliminated on all 17. Calibration: 0.748 is level with the 0.746 of
the untrained network, so this restores the baseline rather than improving on
it. Deep SVDD assumes a clean one-class training set while PyOD fits it
unsupervised on contaminated data; training on genuinely no... (continued)

18834 of 20378 relevant lines covered (92.42%)

10.17 hits per line

Coverage Regressions

Lines Coverage ∆ File
1
97.99
0.16% pyod/models/deep_svdd.py
1
99.34
0.43% pyod/test/test_deepsvdd.py
Jobs
ID Job ID Ran Files Coverage
1 30678244913.1 01 Aug 2026 01:51AM UTC 189
92.41
GitHub Action Run
2 30678244913.2 01 Aug 2026 01:58AM UTC 189
92.41
GitHub Action Run
3 30678244913.3 01 Aug 2026 02:01AM UTC 189
92.42
GitHub Action Run
4 30678244913.4 01 Aug 2026 02:02AM UTC 189
92.42
GitHub Action Run
5 30678244913.5 01 Aug 2026 02:02AM UTC 189
92.42
GitHub Action Run
6 30678244913.6 01 Aug 2026 02:02AM UTC 189
92.42
GitHub Action Run
7 30678244913.7 01 Aug 2026 02:08AM UTC 189
92.42
GitHub Action Run
8 30678244913.8 01 Aug 2026 02:10AM UTC 189
92.42
GitHub Action Run
9 30678244913.9 01 Aug 2026 02:10AM UTC 189
92.41
GitHub Action Run
10 30678244913.10 01 Aug 2026 02:14AM UTC 189
92.42
GitHub Action Run
11 30678244913.11 01 Aug 2026 02:21AM UTC 189
92.42
GitHub Action Run
Source Files on build 30678244913
  • Tree
  • List 189
  • Changed 2
  • Source Changed 0
  • Coverage Changed 2
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • d0d9eee3 on github
  • Prev Build on master (#30639021565)
  • Next Build on development (#30680515234)
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE TRIAL · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc