@a_proca profile picture
Alexandra Proca @a_proca

PhD student @imperialcollege. theoretical neuroscience, deep learning. 🦋aproca.bsky.social

Joined April 2020
  • Tweets 91
  • Following 320
  • Followers 240
36 Photos and videos
Pinned Tweet
Alexandra Proca@a_proca
20 Jun 2025
How do task dynamics impact learning in networks with internal dynamics? Excited to share our ICML Oral paper on learning dynamics in linear RNNs! with @ClementineDomi6 @mpshanahan @PedroMediano openreview.net/forum?id=KGOc…

Learning dynamics in linear recurrent neural networks

Recurrent neural networks (RNNs) are powerful models used widely in both machine learning and neuroscience to learn tasks with temporal dependencies and to model neural dynamics. However, despite...

openreview.net
4
18
98
15,700
Alexandra Proca retweeted
UniReps@unireps
Jun 11
📢Join us for the next UniReps x @ELLISforEurope speaker series event, happening on 17 June at 4:00 PM CEST with @adityagrover_ and @black_samorez!🚀
4
10
1,138
Excited to share our ICLR Oral paper, co-lead with Pratyaksh Sharma, and with @lucas_prie @PedroMediano! We study how feature geometry is shaped by memory demands in RNNs, introducing the concept of temporal superposition. openreview.net/forum?id=7cMz…
4
14
55
4,033
more replies
We induce spatial superposition by using 5D input (A-E) and temporal superposition by increasing k. As memory demand (k) increases, the RNN drops features in favor of representing others for a longer duration. This representational tradeoff leads to a strategy that is all-or-none
1
2
86
In summary, we expand the study of superposition to recurrent architectures, enabling us to study how temporal information acts as a capacity constraint and affects feature geometry. Our work highlights how superposition affects dynamics and is shaped by time.
3
75
Alexandra Proca retweeted
Jamie Simon@learning_mech
Apr 24
1/ Deep learning is going to have a scientific theory. We can see the pieces starting to come together, and it's looking a lot like physics! We're releasing a paper pulling together these emerging threads and giving them a name: learning mechanics. 🔨 arxiv.org/pdf/2604.21691 🔧
54
292
1,510
304,726
Alexandra Proca retweeted
UniReps@unireps
Apr 17
📢Join us for the next UniReps x @ELLISforEurope speaker series happening on 29th April at 4:00 PM CET with @neurosacramento and Paul Friedrich!👾
5
16
1,573
Alexandra Proca retweeted
UniReps@unireps
Feb 17
📢The next UniReps x @ELLISforEurope speaker series event is happening on 26 February at 4:00 PM CET with @orvieto_antonio and @SaraKangaslahti . Don’t miss it!🚀
1
8
24
3,737
Alexandra Proca retweeted
Andrew Saxe@SaxeLab
14 Jul 2025
Excited to share new work @icmlconf by Loek van Rossem exploring the development of computational algorithms in recurrent neural networks. Hear it live tomorrow, Oral 1D, Tues 14 Jul West Exhibition Hall C: icml.cc/virtual/2025/poster/… Paper: openreview.net/forum?id=3go0… (1/11)

2
20
69
5,781
How do task dynamics impact learning in networks with internal dynamics? Excited to share our ICML Oral paper on learning dynamics in linear RNNs! with @ClementineDomi6 @mpshanahan @PedroMediano openreview.net/forum?id=KGOc…

Learning dynamics in linear recurrent neural networks

Recurrent neural networks (RNNs) are powerful models used widely in both machine learning and neuroscience to learn tasks with temporal dependencies and to model neural dynamics. However, despite...

openreview.net
4
18
98
15,700
more replies
Finally, although many results we present are based on SVD, we also derive a form based on an eigendecomposition, allowing for rotational dynamics and to which our framework naturally extends to. We use this to study learning in terms of polar coordinates in the complex plane.
1
1
8
338
In summary, this work provides a novel flexible framework for studying learning in linear RNNs, allowing us to generate new insights into their learning process and the solutions they find, and progress our understanding of cognition in dynamic task settings.
5
197
Load more