1,232 research outputs found
Supporting visual quality assessment with machine learning
No full textObjective metrics for visual quality assessment often base their reliability on the explicit modeling of the highly non-linear behavior of human perception; as a result, they may be complex and computationally expensive. Conversely, machine learning (ML) paradigms allow to tackle the quality assessment task from a different perspective, as the eventual goal is to mimic quality perception instead of designing an explicit model the human visual system. Several studies already proved the ability of ML-based approaches to address visual quality assessment; nevertheless, these paradigms are highly prone to overfitting, and their overall reliability may be questionable. In fact, a prerequisite for successfully using ML in modeling perceptual mechanisms is a profound understanding of the advantages and limitations that characterize learning machines. This paper illustrates and exemplifies the good practices to be followed.Intelligent SystemsElectrical Engineering, Mathematics and Computer Scienc
Soft-Max circuit design with adjustable gain
No full textSoft-Max (SM) operation in vector signal processing usually serves to remap an input distribution within a predetermined range; by a scalar (gain) parameter one can adjust the sharpness of the overall process. Thus, to exhibit practical interest, design approaches to SM circuitry should be consistent and, at the same time, allow dynamic gain control. Therefore, a preliminary analysis applies a power-series expansion to Soft-Max processing and derives both an analytical upper bound to the resulting approximation error, and a convenient mathematical approach to gain control. Theoretical achievements drive the subsequent current-mode circuit design, which yields a modular architecture that enhances overall parallelism. For simplicity, a digital mechanism supports the dynamic gain control in Soft-Max processing, but analogue solutions are also feasible. Simulation results in both static and dynamic tests confirmed the accuracy and effectiveness of the proposed design method. The cell-based circuit architecture sharply reduces VLSI complexity and limits power consumption
Modellistica noise-like coding di memoria associativa: applicazione ad operazioni di riconoscimento
No full text
A Symbolic-Neural classification System assisting the characterization of the Lyme-disease
No full text- …
