Polytechnic University of Turin

About: Polytechnic University of Turin is a education organization based out in Turin, Piemonte, Italy. It is known for research contribution in the topics: Finite element method & Computer science. The organization has 11553 authors who have published 41395 publications receiving 789320 citations. The organization is also known as: POLITO & Politecnico di Torino.

A Single-Layer Frequency-Selective Surface for Ultrawideband Electromagnetic Shielding

Abstract: An efficient approach to achieve the shielding effectiveness (SE) by using a frequency-selective surface (FSS) is presented. This FSS, which consists of cross dipoles and rings printed on the opposite sides of a single-layer FR-4 substrate, exhibits a wide, 7.5-GHz stopband to provide simultaneous shielding in both X- and Ka-bands. Experimental results confirm SE of the prototype over an ultra-wide band with more than 20-dB measured attenuation. The design is compact and suitable to provide shielding against the radiation interference caused by license-free and other radio systems.

Modeling of surface myoelectric signals. I. Model implementation

Abstract: The relationships between the parameters of active motor units (MU's) and the features of surface electromyography (EMG) signals have been investigated using a mathematical model that represents the surface EMG as a summation of contributions from the single muscle fibers. Each MU has parallel fibers uniformly scattered within a cylindrical volume of specified radius embedded in an anisotropic medium. Two action potentials, each modeled as a current tripole, are generated at the neuromuscular junction, propagate in opposite directions and extinguish at the fiber-tendon endings. The neuromuscular junctions and fiber-tendon endings are uniformly scattered within regions of specified width. Muscle fiber conduction velocity and average fiber length to the right and left of the center of the innervation zone are also specified. The signal produced by MU's with different geometries and conduction velocities are superimposed. Monopolar, single differential and double differential signals are computed from electrodes placed in equally spaced locations on the surface of the muscle and are displayed as functions of any of the model's parameters. Spectral and amplitude variables and conduction velocity are estimated from the surface signals and displayed as functions of any of the model's parameters. The influence of fiber-end effects, electrode misalignment, tissue anisotropy, MU's location and geometry are discussed. Part II of this paper will focus on the simulation and interpretation of experimental signals.

Entropy-SGD: Biasing Gradient Descent Into Wide Valleys.

Abstract: This paper proposes a new optimization algorithm called Entropy-SGD for training deep neural networks that is motivated by the local geometry of the energy landscape. Local extrema with low generalization error have a large proportion of almost-zero eigenvalues in the Hessian with very few positive or negative eigenvalues. We leverage upon this observation to construct a local-entropy-based objective function that favors well-generalizable solutions lying in large flat regions of the energy landscape, while avoiding poorly-generalizable solutions located in the sharp valleys. Conceptually, our algorithm resembles two nested loops of SGD where we use Langevin dynamics in the inner loop to compute the gradient of the local entropy before each update of the weights. We show that the new objective has a smoother energy landscape and show improved generalization over SGD using uniform stability, under certain assumptions. Our experiments on convolutional and recurrent networks demonstrate that Entropy-SGD compares favorably to state-of-the-art techniques in terms of generalization error and training time.