The full list of publications, including citations and updated H-index, are available on Google Scholar.
Journal Articles
Overgeneralization in neural networks is addressed by designing a confidence score derived from theoretical insight on features of the data generating distribution implicitly learnt by denoising autoencoders.
We investigate the trade-off between computational efficiency and accuracy of Izhikevich neuron models by numerically quantifying their convergence to provide design guidelines in choosing the limit time steps during the discretization procedure.
We propose a model of repetition suppression in the mammalian neocortex based on a computational model of cortical self-organization, in particular stressing the role of plasticity in long-range cortical lateral inhibitory interactions.
Intraneural stimulation elicits discrimination of textural features by artificial fingertip in intact and amputee humans
C.M. Oddo,
S. Raspopovic,
F. Artoni,
A. Mazzoni,
G. Spigler,
F. Petrini
F. Giambattistelli,
F. Vecchio,
F. Miraglia,
L. Zollo,
G. Di Pino,
D. Camboni,
M.C. Carrozza,
E. Guglielmelli,
P.M. Rossini,
U. Faraguna,
S. Micera
eLife (2016)
PDF /
Journal Link
Experimental study to provide tactile perception to upper-limb amputee subjects via intraneural stimulation, using a model of human mechano-receptors based on the Izhikevich artificial neuron model.
Peer-Reviewed Proceedings
We explore a variant of the Technological Singularity whereby a number of intelligent agents within a simulated environment achieve progress in science and technology that improves the computational power available to the simulation, and consequently the temporal speed-up it can achieve, which in turns allows for even faster technological progress. Such iterative development may quickly lead to extremely large simulated societies experiencing extreme temporal speed-ups.
A wearable haptic feedback system for assisting lower-limb amputees in multiple locomotion tasks
I. Cesini,
G. Spigler,
S. Prasanna,
D. Taxis,
F. Dell'Agnello,
E. Martini,
S. Crea,
N. Vitiello,
A. Mazzoni,
C. M. Oddo
International Symposium on Wearable Robotics, Pia, Italy (2018)
We present a simplified model of human mechano-receptors based on the Izhikevich artificial neuron model, and we show how the spikes-encoded information retains important spatio-temporal features of the perceived surfaces.
Patents
Books
Conference Posters
We quantify the amount of forgetting on sequential Atari games learning by Deep Reinforcement Learning. Specifically, we show that while catastrophic forgetting is found to be highly disruptive in terms of the behavior of an agent, the changes to the parameters of its neural network are not as drastic as it could seem from the large drop in performance. Indeed, it is found that relatively short periods of retraining on previously learnt tasks are often sufficient to quickly recover and improve the lost performance.
Newspaper Articles