ConvergenceWarning

[siebrenf]

maelstrom throws

ConvergenceWarning: Liblinear failed to converge, increase the number of iterations.

Possible sources:

• The warning likely originates from LinearSVR. https://github.com/vanheeringen-lab/gimmemotifs/blob/ecd4806eff4567744016f96b596d72e25a006e4a/gimmemotifs/maelstrom/moap.py#L591
• The location in the full log suggest it might originate from _df_rank_aggregation.

Full log

2022-12-20 10:30:34 - INFO - Starting maelstrom
2022-12-20 10:30:34 - INFO - Input is not mean-centered, setting the mean of all rows to 0.
2022-12-20 10:30:34 - INFO - Use --nocenter if you know what you're doing and want to change this behavior.
2022-12-20 10:30:34 - INFO - Note that if you use count data (ChIP-seq, ATAC-seq) we recommend to first transform your data, for instance using log2(), and to normalize between samples. To create a table suitable for maelstrom you can use the coverage_table script included with GimmeMotifs.
2022-12-20 10:30:41 - INFO - motif scanning (counts)
2022-12-20 10:30:41 - INFO - reading table
2022-12-20 10:30:44 - INFO - setting threshold
2022-12-20 10:30:51 - INFO - creating count table
2022-12-20 10:37:27 - INFO - done
2022-12-20 10:37:27 - INFO - creating dataframe
2022-12-20 10:39:48 - INFO - motif scanning (scores)
2022-12-20 10:39:48 - INFO - reading table
2022-12-20 10:39:56 - INFO - creating score table (z-score, GC%)
2022-12-20 11:03:36 - INFO - done
2022-12-20 11:03:36 - INFO - creating dataframe
2022-12-20 11:28:25 - INFO - Selecting non-redundant motifs
2022-12-20 11:37:08 - INFO - Selected 585 motifs
2022-12-20 11:37:08 - INFO - Motifs: .../nonredundant.motifs.pfm
2022-12-20 11:37:08 - INFO - Factor mappings: .../nonredundant.motifs.motif2factors.txt
2022-12-20 11:37:19 - INFO - Fitting BayesianRidge
100%|██████████| 33/33 [06:14<00:00, 11.34s/it]
2022-12-20 11:43:36 - INFO - Done
2022-12-20 11:43:48 - INFO - Fitting XGBoostRegression
100%|██████████| 33/33 [1:00:42<00:00, 110.38s/it]
2022-12-20 12:44:33 - INFO - Done
2022-12-20 12:44:45 - INFO - Fitting MultiTaskLasso
2022-12-20 16:08:58 - INFO - Done
2022-12-20 16:09:10 - INFO - Fitting SVR
2022-12-20 19:12:30 - INFO - Done
2022-12-20 19:12:30 - INFO - Rank aggregation

.../site-packages/gimmemotifs/__init__.py:24: ConvergenceWarning: Liblinear failed to converge, increase the number of iterations.
  _warn(*args, **kwargs)
[warning repeats 32 more times ]

2022-12-20 19:12:32 - INFO - Correlation
2022-12-20 19:13:16 - INFO - html report
2022-12-20 19:19:12 - INFO - .../gimme.maelstrom.report.html

[siebrenf]

# sklearn: Prefer dual=False when n_samples > n_features.
# X is in shape (n_samples, n_features)
dual = X.shape[0] <= X.shape[1]
logger.debug(f"dual={dual}")

clf = LinearSVR(dual=dual, random_state=self.random_state)