Recent searches

No recent searches

Search options

Only available when logged in.
sigmoid.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
A social space for people researching, working with, or just interested in AI!

Administered by:

Server stats:

596
active users

sigmoid.social: AboutProfiles directoryPrivacy policy

Mastodon: AboutGet the appKeyboard shortcutsView source codev4.4.2

#autoencoders

0 posts0 participants0 posts today
Dennis Alexis Valin Dittrich

When Dimensionality Hurts: The Role of #LLM Embedding Compression for Noisy Regression Tasks d.repec.org/n?u=RePEc:arx:pape
"… suggest that the optimal dimensionality is dependent on the signal-to-noise ratio, exposing the necessity of feature compression in high noise environments. The implication of the result is that researchers should consider the #noise of a task when making decisions about the dimensionality of text.

… findings indicate that sentiment and emotion-based representations do not provide inherent advantages over learned latent features, implying that their previous success in similar tasks may be attributed to #regularisation effects rather than intrinsic informativeness."
#ML #autoencoders #Overfitting

Lou Zonca

New preprint from our group ! 🧠 💻

*Whole-brain modelling of low-dimensional manifold modes reveals organising principle of brain dynamics*
biorxiv.org/content/10.1101/20

bioRxiv · Whole-brain modelling of low-dimensional manifold modes reveals organising principle of brain dynamicsThe revolutionary discovery of resting state networks radically shifted the focus from the role of local regions in cognitive tasks to the ongoing spontaneous dynamics in global networks. Yet, there is a growing realisation that these resting state networks could be a bit like the shadow tracings in Plato’s famous cave, perhaps mere epiphenomena of an underlying hidden space from where these shadows emanate. Here we used deep variational auto-encoders to extract manifolds of low dimensionality from whole-brain dynamics measured with functional magnetic resonance imaging (fMRI). Crucially, we constructed the first dynamical model of the low dimensional manifold modes, i.e., networks of nodes using non-linear oscillators coupled with the effective functional connectivity, taking into account the level of non-equilibrium dynamics quantified by the non-reversibility of the signals. Irrespective of parcellation size, we found an optimal number of roughly ten manifold modes to best describe the whole-brain activity. Importantly, compared to traditional whole-brain modelling using all the nodes in a parcellation, we obtained better results for resting and task activity by modelling the dynamics of the coupled manifold modes. These findings show the key causal role of manifolds as a fundamental organising principle of brain function at the whole-brain scale, providing evidence that networks of brain regions rather than individual brain regions are the key computational engines of the brain. ### Competing Interest Statement The authors have declared no competing interest.
#brain#modeling#autoEncoders
TrendingLive feeds
About