Lucas Lacasa

🚨new preprint !!

How to describe, forecast or control the dynamics of temporal networks? A possible approach includes using fluid mechanical data-driven tools such as #POD and #DMD.

Preprint: arxiv.org/abs/2509.03135

After lunch, we have begun with Lucas Lacasa (@wetuad.bsky.social) who is introducing the use of #federated #learning as implementable solutions for cross-institutional problems. Highlighting industry applications of quantifying #emergence.

@ifisc.uib-csic.es

Leveraging chaos in the training of artificial neural networks Traditional algorithms to optimize artificial neural networks when confronted with a supervised learning task are usually exploitation-type relaxational dynamics such as gradient descent (GD). Here, w...

🚨New preprint !!

Interpreting the training process of a neural network as a temporal network trajectory, we found a regime where such trajectory becomes chaotic. Rather than a nuisance, such chaotic mixing boosts training!

w/ P. Jiménez, @miguelcsoriano.bsky.social

arxiv.org/abs/2506.08523

🚨 Just published !!

How to extract a *scalar time series* that accurately captures the dynamics of a whole temporal network ?

#netsci2025 too bad I missed you.

Great Collab w/ Lluís Arola, Naoki Masuda and F. Javier Marín

Open access --> www.sciencedirect.com/science/arti...

New Study Out Now!
We investigated how deep brain stimulation (DBS) of the nucleus accumbens (NAc) affects memory. www.nature.com/npp/

Title: NAc-DBS selectively enhances memory updating without effect on retrieval
👇 THREAD 👇

🆕 Preprint on the use of Machine Learning tools (deep reinforcement learning + transfer learning) for aerodynamic optimization !!

w/ David Ramos, Gonzalo Rubio & Eusebio Valero

arxiv.org/pdf/2505.02634

joder y yo sin enterarme!! Esos máquinas!!

New paper on Machine Learning in Aerostructures, now published in Expert Systems With Applications:

shorturl.at/Olfig

Great industry-academia collaboration
@aeroespacialupm.bsky.social
@esa.int
@ifisc.uib-csic.es @EsAirbus

We have looked at how 🧠evolves in graph space. We found that their network trajectories characterize well aging and brain pathologies. Details 👇

Today, we are publishing the first-ever International AI Safety Report, backed by 30 countries and the OECD, UN, and EU.

It summarises the state of the science on AI capabilities and risks, and how to mitigate those risks. 🧵

Full Report: assets.publishing.service.gov.uk/media/679a0c...

1/21

How to describe the dynamical properties of networks *with no labels* that change over time ?

🆕 preprint
(with @_CaligiuriLisa_ @tobiasgalla.bsky.social)

arxiv.org/abs/2412.14864

And the other way around!

IMHO confusion for neuroscientists come from the widespread Hebbian motto causing conflation.

Anyway, the crucial difference is whether information can propagate between two “connected” nodes thanks to the presence of such link. If yes, then it is a connection, if no, at most it is a relation.

So @naturecomms.bsky.social has highlighted our research home @ifisc.uib-csic.es as an institutional model for interdisciplinary science!

🚨 New preprint out!

We build **scalar** time series embeddings of temporal networks !

The key enabling insight : the relevant feature of each network snapshot... is just its distance to every other snapshot!

Work w/ FJ Marín, N. Masuda, L. Arola-Fernández

arxiv.org/abs/2412.02715

Right on top of my to-read pile !! Very timely as we’re about to post a preprint on TN embedding :-)

A thread is available in the other blue place twitter.com/wetuad/statu...

Hi Bluesky! My first skytweet is to reflect on how more is different when individual AIs come together arxiv.org/abs/2310.12802 #complexsystems #emergence #machinelearning