[Eeglablist] Practically reproducible results with AMICA

Tue Oct 6 12:13:32 PDT 2026

Grace -

You are working in a fruitful - and to date too little explored -
direction, I believe.
Have you looked into Fiorenzo Artoni's RELICA
<https://urldefense.com/v3/__https://www.sciencedirect.com/science/article/pii/S1053811914007526?casa_token=5ekIDTOt9FoAAAAA:uYbo9CtmaBau_DYN5dbNMNGLiLXzVl1f0omPA4YfwHKEzlhpdjvzyg-tfAVKnmE8EvSFtBuPHA__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvT1hy3qM$ >(=
'reliability of ICA components') (currently, with 123 Google Scholar
citations)?

In the last version of RELICA I am aware of, we dexplictly defined two ICA
decompositions - one as the standard and one to support bootstrap
replications (the same number of data points, from the same dataset, only
pulled at random WITH replacement (so one point might be selected > 1
times). This method has statistical support - it creates variation while
never moving away from the underlying data point (scalp map) distribution.
Now, Jason's latest compile of AMICA is quite fast (especially on current
machines), which should allow using AMICA for both the standard and
bootstrap decompositions.

Fiorenzo's strategy in RELICA is to cluster all the resulting components -
to see which show up invariably, and which do less so...
(You might want to save and/or plot - for each resulting IC cluster - pdfs
of exactly WHICH data points are used in the cluster decompositions, as it
might be that some sources only show up in some portion(s) of the data. I
believe source space nonstationarity will be the next frontier in EEG
analysis).

Now, with Gwen (see References below) we recently set out to determine the
value of running more than the (heuristic) default 2,000 iterations in
AMICA decompositions. Results showed that yes, the resulting ICs do
slightly more independent (and have more simple, 'dipolar' scalp maps) the
more iterations you let AMICA run.  But at 2,000 iterations, further
component changes are relatively small, and quite slow to appear. This
should also depend to some extent on data length and on channel number, of
course - not something we had the (computer or time) bandwidth to explore.

Feel free to write me if you are tempted to try any of this, as I might
have some further thoughts, ...

 [PDF] ieee.org
<https://urldefense.com/v3/__https://ieeexplore.ieee.org/iel8/11588304/11588911/11588993.pdf?casa_token=DkA0oYZeojQAAAAA:RbmpY9HpACIT8EDeV1UJjyvY7OOp54KqHP7zsxkVpgFxbuPodk7nvLrjqizl7gLYEgMc49fztw__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvd_4fewB$ >
Quantitative Exploration of Sufficient Data  Quantities for EEG
Decomposition Using *AMICA*
<https://urldefense.com/v3/__https://ieeexplore.ieee.org/abstract/document/11588993/?casa_token=Ms2McBi_P4QAAAAA:TjnwsPctnf_Sg5kZpZVMAV3seVETLVemHIBt4US4lRrgIGTF4OneHK9JV3XlB2TkFcj04wpzfQ__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvb5BTm3g$ >
G Frank
<https://urldefense.com/v3/__https://scholar.google.com/citations?user=ZMFrRcsAAAAJ&hl=en&oi=sra__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvefBlfn2$ >, SY
*Shirazi*
<https://urldefense.com/v3/__https://scholar.google.com/citations?user=9QzpcEoAAAAJ&hl=en&oi=sra__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvUTOrvgm$ >, J
*Palmer*
<https://urldefense.com/v3/__https://scholar.google.com/citations?user=coJRC_8AAAAJ&hl=en&oi=sra__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvavAwGTr$ >… -
2025 12th …, 2025 - ieeexplore.ieee.org
… Unless mentioned otherwise throughout this paper *AMICA* is run with a
maximum of
3000 … [10] JA *Palmer*, K. Kreutz-Delgado, and S. *Makeig*, “*Amica*: An
adaptive mixture of …
Save Cite Related articles
<https://urldefense.com/v3/__https://scholar.google.com/scholar?q=related:LrEIpEcpAOMJ:scholar.google.com*&scioq=palmer*makeig*shirazi*AMICA&hl=en&as_sdt=0,34__;LysrKw!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvYyWzeDp$ >
[PDF] ieee.org
<https://urldefense.com/v3/__https://ieeexplore.ieee.org/iel7/10431814/10431802/10431896.pdf?casa_token=S2CitxisetQAAAAA:bSVkVy00imy5Mzcz0nWsBw0RqJooBffAnlXZHVYhFVDYZbP1AS43Tg6dbm5siquhjglLwSz9Pg__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvdVcu-ji$ >
An exploration of optimal parameters for efficient blind source separation
of EEG recordings using *AMICA*
<https://urldefense.com/v3/__https://ieeexplore.ieee.org/abstract/document/10431896/?casa_token=8wECD0GXuXUAAAAA:1wAkBCqqvsKpOCcr2qzVtauRWK_R_wcMR1Ut-pn8cGXyp_Q5URzSY6Is2F3UHv3gprXxgX01hQ__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvVUtGDeJ$ >
G Frank
<https://urldefense.com/v3/__https://scholar.google.com/citations?user=ZMFrRcsAAAAJ&hl=en&oi=sra__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvefBlfn2$ >, SY
*Shirazi*
<https://urldefense.com/v3/__https://scholar.google.com/citations?user=9QzpcEoAAAAJ&hl=en&oi=sra__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvUTOrvgm$ >, J
*Palmer*
<https://urldefense.com/v3/__https://scholar.google.com/citations?user=coJRC_8AAAAJ&hl=en&oi=sra__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvavAwGTr$ >… -
2023 IEEE 23rd …, 2023 - ieeexplore.ieee.org
… *AMICA* is considered the benchmark against which to compare the
performance of other ICA
algorithms for EEG decomposition. *AMICA* … Here, 70-channel *AMICA*
decompositions are …
Save Cite Cited by 6
<https://urldefense.com/v3/__https://scholar.google.com/scholar?cites=9187053168422081348&as_sdt=5,34&sciodt=0,34&hl=en__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvWY240LF$ >
Related
articles
<https://urldefense.com/v3/__https://scholar.google.com/scholar?q=related:RNNsfC74fn8J:scholar.google.com*&scioq=palmer*makeig*shirazi*AMICA&hl=en&as_sdt=0,34__;LysrKw!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvTBDq3sx$ >
All
5 versions
<https://urldefense.com/v3/__https://scholar.google.com/scholar?cluster=9187053168422081348&hl=en&as_sdt=0,34__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvfoja929$ >
[PDF] arxiv.org <https://urldefense.com/v3/__https://arxiv.org/pdf/2506.10156__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvbmv0HMW$ >
Quantifying data requirements for EEG independent component analysis using
*AMICA* <https://urldefense.com/v3/__https://arxiv.org/abs/2506.10156__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvTBWzLf6$ >
G Frank
<https://urldefense.com/v3/__https://scholar.google.com/citations?user=ZMFrRcsAAAAJ&hl=en&oi=sra__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvefBlfn2$ >, SY
*Shirazi*
<https://urldefense.com/v3/__https://scholar.google.com/citations?user=9QzpcEoAAAAJ&hl=en&oi=sra__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvUTOrvgm$ >, J
*Palmer*
<https://urldefense.com/v3/__https://scholar.google.com/citations?user=coJRC_8AAAAJ&hl=en&oi=sra__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvavAwGTr$ >… -
arXiv preprint arXiv …, 2025 - arxiv.org
… Additionally, *AMICA* can learn stationary subsets of data and fit …
Unless mentioned otherwise
throughout this paper *AMICA* is … *AMICA* was selected for this study
because prior work has …
Save Cite Cited by 2
<https://urldefense.com/v3/__https://scholar.google.com/scholar?cites=15454239581601227015&as_sdt=5,34&sciodt=0,34&hl=en__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvQSaK9Ge$ >
Related
articles
<https://urldefense.com/v3/__https://scholar.google.com/scholar?q=related:B2Ef9wR9eNYJ:scholar.google.com*&scioq=palmer*makeig*shirazi*AMICA&hl=en&as_sdt=0,34__;LysrKw!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvb8UyqC8$ >
All
2 versions
<https://urldefense.com/v3/__https://scholar.google.com/scholar?cluster=15454239581601227015&hl=en&as_sdt=0,34__;!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvQt953O5$ >

On Tue, Oct 6, 2026 at 11:59 AM Grace Harvie via eeglablist <
eeglablist at sccn.ucsd.edu> wrote:

> Hi All,
>
> I'm back with another AMICA question - this time regarding obtaining
> practically reproducible decompositions with AMICA.
>
> I was able to obtain consistent and reproducible results with extended
> INFOMAX by utilising the parameters described by Makoto's here:
> https://eeglab.ucsd.edu/wiki/Makoto%27s_useful_EEGLAB_code#How_to_obtain_practically_reproducible_ICA_results_(09/26/2022_added,_05/08/2024_updated) 
>
> While Makoto comments that the default parameters for AMICA are much more
> refined and seem to produce similarly stable, replicable results, I have
> observed variable results over multiple runs with AMICA - please see the
> top 7 ICs for ten AMICA runs on the same dataset here:
> https://urldefense.com/v3/__https://github.com/graematterneuro/ACTIONStudyEEGData/blob/main/AMICADecompositionReproducibility.png__;!!Mih3wA!ECb3eiORNCeyUP1u_NEtlhsE66TmRqljEcyRtGZyp6WqAlxLU16gwSe69jtFfeZAUuOzjacBKM-x3Q40quQJDJc52OiFBKxr$
>
> Additionally, on one run (Run 8), I received a warning about the ICA
> activities and weight matrix not matching as soon as AMICA finished. I'm
> not sure why I got that warning, especially since EEG.icaact and
> EEG.icaweights data structures appear to each have the correct number of
> values. I'm also not sure if it's relevant to the issue at hand, but I
> thought it was probably a good idea to mention it anyway.
>
> Since Makoto showed that INFOMAX reproducibility could be improved by
> increasing the number of iterations and lowering the learning rate, I tried
> to similarly tweak these parameters in AMICA, however it didn't seem to
> help.
>
> Does anyone have any recommendations to help improve reproducibility of
> AMICA results?
>
> Many thanks,
> Grace Harvie
>
> PhD Candidate, Faculty of Medicine and Health, University of Sydney
>
> Research Assistant, Brain Dynamics Centre, Westmead Institute for Medical
> Research
>
> 176 Hawkesbury Road
> <https://urldefense.com/v3/__https://www.google.com/maps/search/176*Hawkesbury*Road?entry=gmail&source=g__;Kys!!Mih3wA!AKnz6iSru24yz9DWj5ZO1EEELTPJ_TxTdxH99rRiuAbWmPtLxpWg2UiCQYQrozNN-j54SRso-Q2xvS8DXgE5$ >
> | PO Box 412 | Westmead NSW 2145 Australia
> ghar8070 at uni.sydney.edu.au
> westmeadinstitute.org.au
>
>
> This email is private and confidential to the intended recipient. If you
> are not the intended recipient, please do not copy it, circulate it or take
> any action in reliance on it. Kindly notify me that it has been misdirected
> and then delete it. Thank you.
>
> _______________________________________________
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu or visit
> https://sccn.ucsd.edu/mailman/listinfo/eeglablist .
>

-- 
Scott Makeig, Research Scientist and Director, Swartz Center for
Computational Neuroscience, Institute for Neural Computation, University of
California San Diego, La Jolla CA 92093-0559, http://sccn.ucsd.edu/~scott 