Paul Albert

New paper at ICLR 25 ! arxiv.org/abs/2502.00987 github.com/PaulAlbert31/R… We evidence the low-rank bottleneck of LoRA, showing that scaling trainable parameters isn't always enough. Plus, RandLoRA optimizes gradient computation to reduce memory use over LoRA ! #peft #ICLR2025

@TheAIML Co-authors: Dr. @fredzzhang Dr. Hemanth Saratachandran Dr. Cristian Rodriguez-Opazo Dr. Ehsan Abbasnejad and Prof @AntonvdH

This research was conducted in collaboration with co-authors at @TheAIML under the CAR program. We invite you to explore the paper and accompanying codebase: arxiv.org/pdf/2502.00987

New paper at ICLR 25 ! arxiv.org/abs/2502.00987 github.com/PaulAlbert31/R… We evidence the low-rank bottleneck of LoRA, showing that scaling trainable parameters isn't always enough. Plus, RandLoRA optimizes gradient computation to reduce memory use over LoRA ! #peft #ICLR2025

Check out our new work freshly accepted at #neurips2024 where we improve knowledge composition and few-shot generalization in large transformer models using learnt anisotropic scaling of task vectors. Paper: arxiv.org/abs/2407.02880 Code: github.com/fredzzhang/atl…

Excited to announce that the codebase for our #neurips2024 paper Knowledge Composition using Task Vectors with Learned Anisotropic Scaling (arxiv.org/abs/2407.02880) is now available. In this paper, we showed how to combine or transfer the knowledge across different models. [1/6]

Co-authors: @jvlmdr @ArazoEric @tkrishnna @oconnorn

#ECCV2024 is on ! Check out our work on learning from noisy datasets where we observe that detecting out-of-distribution training images using unsupervised learning discards highly important clean examples and how to recover them. @insight_centre @TheAIML eccv.ecva.net/virtual/2024/p…

Congrats to @PaulAlbert31 who successfully defended his PhD thesis "Deep Learning for Computer Vision constrained by Limited Supervision". Jointly-supervised with @kevinmcguinness. Thanks to examiners @fosterjendublin & Prof Yannis Patras of @QMEECS @VistaMilk @insight_centre

Our label noise robust algorithm PLS detects incorrect true label guesses using our pseudo-loss to combat label noise in image classification datasets arxiv.org/abs/2210.04578. Co-authors: @ArazoEric @tkrishnna @oconnorn @kevinmcguinness

The end of a great poster session at #WACV2023 in sunny Waikoloa Hawaii ! I presented PLS my latest label noise paper . This is the result of a great ongoing collaboration with the support of @VistaMilk and @insight_centre. Co-authors 👇

Congratulations to our researchers Paul Albert and Hannah Masterson on winning awards for best oral presentation at the VistaMilk industry day 👏 @scienceirel @agriculture_ie

The code for our latest label noise paper PLS arxiv.org/abs/2210.04578 was just released on github github.com/PaulAlbert31/P…. Feel free to reproduce our results and test on your own noisy dataset ! Team: @ArazoEric @tkrishnna @oconnorn @kevinmcguinness

Just out of an intense poster session #ECCV2022 on Tel Aviv. We detect two forms of label noise using unsupervised contrastive leaning. Paper: arxiv.org/abs/2207.01573. A fruitful collaboration between @insight_centre and @VistaMilk funded by @scienceirel

I just presented my latest work on computer vision in the presence of label noise to the Insight Smart Surrounding showcase #InsightSSS22 @DCU. A research collaboration between by @VistaMilk and @insight_centre @scienceirel. Great to see everyone in person !

Our paper: "Is your noise correction noisy? PLS: Robustness to label noise with two stage detection" was accepted at #WACV2023 ! We detect incorrect true label guesses using our pseudo-loss to combat label noise in image classification datasets arxiv.org/abs/2210.04578