Manlio De Domenico · @manlius
2130 followers · 596 posts · Server mathstodon.xyz
Manlio De Domenico · @manlius
2110 followers · 579 posts · Server mathstodon.xyz

A new issue (n° 9) of is out!

manlius.substack.com/p/complex

Spread the word and suggest a Thought for the next time ;)

_________

Not yet subscribed? It's free, check it out: manlius.substack.com/

Want to suggest a Thought to cover for our “From the community”? Use this link: forms.gle/FW4EHEmfWYyXj3TFA

@networkscience @complexsystems @neuroscience

#datascience #networkscience #bioinspiredcomputing #complexsystems #ComplexityThoughts

Last updated 2 years ago

Manlio De Domenico · @manlius
2076 followers · 577 posts · Server mathstodon.xyz
Luis M. Rocha · @lmrocha
164 followers · 146 posts · Server qoto.org
Luis M. Rocha · @lmrocha
160 followers · 129 posts · Server qoto.org
Manlio De Domenico · @manlius
2027 followers · 556 posts · Server mathstodon.xyz

Are you ready for the Issue #6?

From network foundations to global systems, embodiment and network medicine.

Not yet subscribed? It's free, check it out: manlius.substack.com/

@networkscience @complexsystems @neuroscience

#NetworkMedicine #bioinspiredcomputing #complexsystems #ComplexityThoughts

Last updated 2 years ago

Manlio De Domenico · @manlius
2027 followers · 556 posts · Server mathstodon.xyz

Do we need to hardwire hierarchical connectivity in recurrent neural networks to achieve predictive coding (PC)?

PC is an emergent consequence of an energy efficiency principle, leading to separate subpopulations of prediction and error units.

cell.com/patterns/fulltext/S26

@neuroscience@a.gup.pe @networkscience@a.gup.pe @complexsystems

@PessoaBrain @WiringtheBrain @kordinglab @ricard_sole @albertcardona @c4computation

#computationalbiology #bioinspiredcomputing #NetworkNeuroscience #complexsystems

Last updated 2 years ago

Manlio De Domenico · @manlius
2008 followers · 540 posts · Server mathstodon.xyz

“intrinsic stochasticity is probably positively utilized by brains, not only to optimize the signal response but also to induce intelligence itself, and one of the key roles, played by astrocytes in information processing, could be dealing with noises”

@neuroscience

journals.aps.org/prresearch/ab

#machinelearning #computationalbiology #bioinspiredcomputing

Last updated 2 years ago

Manlio De Domenico · @manlius
2008 followers · 539 posts · Server mathstodon.xyz

The authors propose a simple single-neuron model where the dendrite is implemented as a sequence of thresholded linear units. They manipulate the underlying architecture and investigate the impacts of binary branching constraints and repetition of synaptic inputs on neural computation, finding that models with such manipulations can perform (surprisingly?) well on machine learning tasks.

@neuroscience

direct.mit.edu/neco/article-ab

#machinelearning #bioinspiredcomputing #neuroscience

Last updated 2 years ago

Manlio De Domenico · @manlius
1982 followers · 531 posts · Server mathstodon.xyz
Paul Marrow · @evopma
0 followers · 1 posts · Server ecoevo.social
Manlio De Domenico · @manlius
780 followers · 252 posts · Server mathstodon.xyz

“Neuroevolution enables important capabilities that are typically unavailable to gradient-based approaches, including learning neural network building blocks (for example activation functions), hyperparameters, architectures and even the algorithms for learning themselves [...] This Review looks at several key aspects of modern neuroevolution”

nature.com/articles/s42256-018

#bioinspiredcomputing #ann #neuroevolution #machinlearning #neuroscience

Last updated 2 years ago

Manlio De Domenico · @manlius
780 followers · 252 posts · Server mathstodon.xyz

There are some limitations in our study, but we think it's a promising step towards a better understanding of the relationship between structure and computing, ubiquitous features of empirical complex adaptive systems.

/end

Of interest for:

#neuroscience #computationalbiology #bioinspiredcomputing #statisticalphysics #CollectiveBehavior #ComplexNetworks #complexsystems

Last updated 2 years ago

Manlio De Domenico · @manlius
780 followers · 252 posts · Server mathstodon.xyz

It could be computationally explained by an overfitting effect. It also agrees w/ some literature according to which biological networks have optimized topologies capable of suppressing noisy fluctuations that consist in sparser and more hierarchically organized structures

9/

Of interest for:

#neuroscience #computationalbiology #bioinspiredcomputing #statisticalphysics #CollectiveBehavior #ComplexNetworks #complexsystems

Last updated 2 years ago

Manlio De Domenico · @manlius
780 followers · 252 posts · Server mathstodon.xyz

If we run our analysis for different input signals, we find that when the signals that have to be predicted are noisy the performances for reservoirs w/ modular structures are lower than the one evaluated with empirical connectomes, which is puzzling.

8/

Of interest for:

#neuroscience #computationalbiology #bioinspiredcomputing #statisticalphysics #CollectiveBehavior #ComplexNetworks #complexsystems

Last updated 2 years ago

Manlio De Domenico · @manlius
780 followers · 252 posts · Server mathstodon.xyz

Comparing the outcomes of the corresponding generative models, we find that the number of independent activation states increases with the complexity of the model and organism brain volume, the topology playing an important role w/ its link density and modularity.

7/

Of interest for:

#neuroscience #computationalbiology #bioinspiredcomputing #statisticalphysics #CollectiveBehavior #ComplexNetworks #complexsystems

Last updated 2 years ago

Manlio De Domenico · @manlius
780 followers · 252 posts · Server mathstodon.xyz

We find that system's size (ie the number of neurons in the reservoir) and the nature of input signal have a strong impact. Stochastic and deterministic signals lead to qualitative & measurable differences in prediction.

6/

Of interest for:

#neuroscience #computationalbiology #bioinspiredcomputing #statisticalphysics #CollectiveBehavior #ComplexNetworks #complexsystems

Last updated 2 years ago

Manlio De Domenico · @manlius
780 followers · 252 posts · Server mathstodon.xyz

Our setup:

- use different types of input signals (deterministic chaotic and (un)correlated stochastic)
- use 45 empirical connectomes as reservoirs
- use generative models from different statistical ensemble to control which feature is relevant for a predictive task

5/

Of interest for:

#neuroscience #computationalbiology #bioinspiredcomputing #statisticalphysics #CollectiveBehavior #ComplexNetworks #complexsystems

Last updated 2 years ago

Manlio De Domenico · @manlius
780 followers · 252 posts · Server mathstodon.xyz

Usually, the reservoir is a random network initialized w/ random weights.

We asked: what happens if we put structured networks with controllable modularity, heterogeneity, etc? We have found that a lot can happen & the final performance does not depend trivially on topology

4/

Of interest for:

#neuroscience #computationalbiology #bioinspiredcomputing #statisticalphysics #CollectiveBehavior #ComplexNetworks #complexsystems

Last updated 2 years ago

Manlio De Domenico · @manlius
780 followers · 252 posts · Server mathstodon.xyz

In practice, it allows to map a complex signal into a high-dimensional space where the underlying attractor is reconstructed and then used to forecast the dynamics in the next future, with amazing performances.

3/

Of interest for:

#neuroscience #computationalbiology #bioinspiredcomputing #statisticalphysics #CollectiveBehavior #ComplexNetworks #complexsystems

Last updated 2 years ago