Míriam Barrabés i Torrella

Researcher @Stanford / ML Engineer @GalateaBio. Passionately building AI for positive impact.

[Míriam Barrabés i Torrella @BarrabesMiriam]

🚀 Exciting news! 🚀 The code for DataFix, our latest NeurIPS paper focusing on shift detection and correction, is now available on GitHub!

https://t.co/xRIJfDY1Ei

@_danielmas @ritageleta @DocXavi @alexGioannidis

[Alexander G. Ioannidis @alexGioannidis]

Lots of standardized benchmarks.

Phase error correction for LAI, the only scalable method with this.

Insights...including that Native American dogs 🐕 did not vanish; living on (in part) in the Mexican Xoloitzcuintli. Also, Chihuahuas have a bit of coyote in them.

Credit also to contributor,
@Rich_Rast
And maestro,
@_danielmas

Use it with our toolkit,
https://t.co/bRBktspSzP

[Míriam Barrabés i Torrella @BarrabesMiriam]

Really exciting to see Gnomix finally out!

[Alexander G. Ioannidis @alexGioannidis]

Gnomix is published!

https://t.co/8wq4951VCF

Still the most accurate & fastest local ancestry inference (LAI) method, bests Flare and others (Recomb-Mix & Orchestra) released since our preprint.

Easy use
https://t.co/RoAzHobn7e

Leads @HelgiHilmars @arvind0422 @BarrabesMiriam

[Míriam Barrabés i Torrella @BarrabesMiriam]

Happy to see our work getting attention 🚀. Feedback and contributions are very welcome!

[Stephen Turner 🦋 @stephenturner.us @strnr]

snputils (https://t.co/pyf7FyY1dF): A High-Performance Python Library for Genetic Variation and Population Structure https://t.co/wEi2mHYFbA 🧬💻🧪 https://t.co/bJALQDZpV2

[Daniel Mas Montserrat @_danielmas]

Excited to share our preprint on iLTM: an Integrated Large Tabular Model!

arxiv: https://t.co/ZpT5ZiKJs9

No single technique consistently excels across all tabular tasks. iLTM addresses this by integrating distinct paradigms in a single architecture:

- Gradient Boosted Decision Trees (GBDT)-based embeddings
- Robust preprocessing with random feature projections
- A meta-trained hypernetwork
- Retrieval augmentation with Soft Nearest Neighbors

@d_bonet @marcal_cc @alexGioannidis
(1/N)

[Statistics (Machine Learning) Papers @StatsPapers]

Feature Shift Localization Network. https://t.co/znN07E2vuq

[Galatea Bio Inc. @GalateaBio]

We’re excited to be part of #PMWC2025 —Our session: “Powering Discoveries in Medicine by Pairing Genetic Diversity with AI-enabled Data Analytics” is part of Track B and will be delivered by our CEO, Carlos D. Bustamante on February 7th at 10:15am PT.

https://t.co/J6Q2N7cBCD

[Míriam Barrabés i Torrella @BarrabesMiriam]

Great work!! 👏👏👏

[Arturo @arturolp]

Happy to share our latest paper on improving performance of PRS models across various ethnic groups. #PolygenicRiskScore #PRS @GalateaBio @genomelink @SoniaMorenoGrau @cdbustamante @HumanGenomics @PGSCatalog
https://t.co/XKGudGJBqC

[Arturo @arturolp]

Excited to share our latest PRS work! Our @GalateaBio and @genomelink team performed a comprehensive analysis of published @PGSCatalog models along with locally trained models using LDPred2, PRS-CSx, and SNPnet, across diverse populations using @UKBIOBANK and our own data

[Daniel Mas Montserrat @_danielmas]

Introducing "HyperFast: Instant Classification for Tabular Data" at @RealAAAI, which received the Best Paper Award at @NeurIPSConf Table rep. workshop @TrlWorkshop!

We provide easy-to-use sklearn-like code: https://t.co/qrMF6XStAA

Some insights of the work below 👇🧵(1/N)

[Daniel Mas Montserrat @_danielmas]

Check this awesome talk by @ritageleta at @dlbcnai about our #NeurIPS2023 paper with @BarrabesMiriam @DocXavi @alexGioannidis !

https://t.co/aRR5hy1UCC

[Míriam Barrabés i Torrella @BarrabesMiriam]

Loving the annual reunion at the Deep Learning Barcelona Symposium! Grateful to everyone who expressed interest in our work on homogenizing genomic databases 🧬. And a special shoutout to @ritageleta for delivering an excellent talk on our DataFix tool (NeurIPS) #DLBCN #Catalonia

[Alexander G. Ioannidis @alexGioannidis]

Also, congrats🙌 to another amazing student @BarrabesMiriam together w/ @_danielmas, @ritageleta & @DocXavi, whose DataFix system for feature shift detection, inspired by merging genomic datasets, but much more general, was just presented at #NeurIPS2023.

https://t.co/Lkktywsfca

[Daniel Mas Montserrat @_danielmas]

We presented our paper paper "Adversarial Learning for Feature Shift Detection and Correction" at @NeurIPSConf

We want to thank @ykilcher for comming by our poster!

https://t.co/hidnpTpTy4

#NeurIPS #NeurIPS2023 #AI

[Daniel Mas Montserrat @_danielmas]

Combining data from many sources to increase the sample size is a common step in biomedical studies.

Merging datasets typically involves domain-specific heuristics. But, how can you evaluate the quality of the final dataset?

DataFix can localize and correct potential errors!

[Daniel Mas Montserrat @_danielmas]

We will be presenting the paper "Adversarial Learning for Feature Shift Detection and Localization" at @NeurIPSConf next week! Work done with @BarrabesMiriam @ritageleta @DocXavi @alexGioannidis #NeurIPS2023

[Stat.ML Papers @StatMLPapers]

Adversarial Learning for Feature Shift Detection and Correction. (arXiv:2312.04546v1 [cs.LG]) https://t.co/AwwOnO5B5O

[Daniel Mas Montserrat @_danielmas]

The work combines iterative heuristics with Random Forest and Boosting Trees in order to localize and correct the features that suffer a distribution shift providing state-of-the-art performance.

@NeurIPSConf #NeurIPS #NeurIPS2023 #AI
(3/3)

[Daniel Mas Montserrat @_danielmas]

Our method DataFix for shift detection and correction is now available on GitHub:
https://t.co/T7DclMZymC

DataFix is presented in our paper "Adversarial Learning for Feature Shift Detection and Correction" at @NeurIPSConf #NeurIPS #NeurIPS2023 #AI

(1/3)

[Daniel Mas Montserrat @_danielmas]

The paper "Adversarial Learning for Feature Shift Detection and Correction" was co-lead by @BarrabesMiriam with the excellent help of @ritageleta @DocXavi @alexGioannidis

DataFix provides an effective way to localize and correct corrupted features within a dataset.

(2/3)