1,720,991 research outputs found
An explainable fast deep neural network for emotion recognition
Get PDFIn the context of artificial intelligence, the inherent human attribute of engaging in logical reasoning to
facilitate decision-making is mirrored by the concept of explainability, which pertains to the ability of
a model to provide a clear and interpretable account of how it arrived at a particular outcome. This
study explores explainability techniques for binary deep neural architectures in the framework of emotion
classification through video analysis. We investigate the optimization of input features to binary classifiers for
emotion recognition, with face landmarks detection, using an improved version of the Integrated Gradients
explainability method. The main contribution of this paper consists of the employment of an innovative
explainable artificial intelligence algorithm to understand the crucial facial landmarks movements typical
of emotional feeling, using this information for improving the performance of deep learning-based emotion
classifiers. By means of explainability, we can optimize the number and the position of the facial landmarks
used as input features for facial emotion recognition, lowering the impact of noisy landmarks and thus
increasing the accuracy of the developed models. To test the effectiveness of the proposed approach, we
considered a set of deep binary models for emotion classification, trained initially with a complete set of
facial landmarks, which are progressively reduced basing the decision on a suitable optimization procedure.
The obtained results prove the robustness of the proposed explainable approach in terms of understanding the
relevance of the different facial points for the different emotions, improving the classification accuracy and
diminishing the computational cost
Design of an LSTM cell on a quantum hardware
No full textThe present paper proposes a methodology to implement a Long Short-Term Memory cell in the quantum framework, where inference is computed by replicating the internal structure of the cell using quantum circuits. A suitable encoding is proposed and the design of each quantum operation is detailed. A complexity analysis of the circuit is hence conducted and finally, the quantum architecture is experimentally validated both in an IBM Q simulator and with a numerical simulation on a classical device. The proposed approach leads the way for a completely quantum implementation of a Long Short-Term Memory network
A distributed algorithm for the cooperative prediction of power production in PV plants
No full textForecasting the energy production of photovoltaic plants is today an essential tool for asset owners because it has direct economic implications on the net operating income of the plants whose generated energy is sold in competitive electricity markets. In this paper, we propose an innovative distributed decentralized prediction technique for the forecasting of power generated by several PV plants. The prediction technique is based on neural Echo State Networks (ESN), that are very promising in terms of prediction performance and model accuracy, in conjunction with a distributed learning algorithm and a distributed consensus protocol that makes unnecessary any central coordinator. The technique has been properly conceived for asset owners that hold a large portfolio of PV systems that are geographically spread out over large areas. The algorithm reveals very efficient, accurate, scalable to any number of plants and it requires only a simple reliable communication channel. Through a real-world example we assess the applicability of such methodology and show its strength
Water quality prediction based on wavelet neural networks and remote sensing
No full textWavelet artificial neural networks and remote sensing techniques can be used to estimate water quality variables such as Chlorophyll-a, turbidity and suspended solids. This paper describes empirical algorithms for the estimation of these variables incorporating information from the Operational Land Imager Sensor on board the Landsat-8 satellite.
Neural networks are seasonally trained using data from the Cefni reservoir (Anglesey, U.K.), covering a variety of physical trophic status. Chlorophyll-a levels and the suspended solids are estimated from the reflectance in band-2 and band-4, while the turbidity values from reflectance in band-4. Experimental results show the potential of Landsat-8 as a substitute of Landsat-7 in water bodies quality monitoring. Moreover, predicted values obtained by using wavelet artificial neural networks fit well measured data and hence, such models provide accurate results therefore improving the efficiency in monitoring water quality parameters and contributing to possible decision making processes in the environmental management
Deep Neural Networks for Multivariate Prediction of Photovoltaic Power Time Series
Get PDFThe large-scale penetration of renewable energy sources is forcing the transition towards
the future electricity networks modeled on the smart grid paradigm, where energy clusters call for new
methodologies for the dynamic energy management of distributed energy resources and foster to form
partnerships and overcome integration barriers. The prediction of energy production of renewable energy
sources, in particular photovoltaic plants that suffer from being highly intermittent, is a fundamental tool
in the modern management of electrical grids shifting from reactive to proactive, with also the help of
advanced monitoring systems, data analytics and advanced demand side management programs. The gradual
move towards a smart grid environment impacts not only the operating control/management of the grid, but
also the electricity market. The focus of this article is on advanced methods for predicting photovoltaic
energy output that prove, through their accuracy and robustness, to be useful tools for an efficient system
management, even at prosumer's level and for improving the resilience of smart grids. Four different deep
neural models for the multivariate prediction of energy time series are proposed; all of them are based on the
Long Short-Term Memory network, which is a type of recurrent neural network able to deal with long-term
dependencies. Additionally, two of these models also use Convolutional Neural Networks to obtain higher
levels of abstraction, since they allow to combine and filter different time series considering all the available
information. The proposed models are applied to real-world energy problems to assess their performance
and they are compared with respect to the classic univariate approach that is used as a reference benchmark.
The significance of this work is to show that, once trained, the proposed deep neural networks ensure their
applicability in real online scenarios characterized by high variability of data, without requiring retraining
and end-user's tricks
Bimodal Feature Analysis with Deep Learning for Autism Spectrum Disorder Detection
No full textAutism Spectrum Disorder (ASD) is a complex and heterogeneous neurodevelopmental disorder which affects a significant proportion of the population, with estimates suggesting that about 1 in 100 children worldwide are affected by ASD. This study introduces a new Deep Neural Network for identifying ASD in children through gait analysis, using features extracted from frames composing video recordings of their walking patterns. The innovative method presented herein is based on imagery and combines gait analysis and deep learning, offering a noninvasive and objective assessment of neurodevelopmental disorders while delivering high accuracy in ASD detection. Our model proposes a bimodal approach based on the concatenation of two distinct Convolutional Neural Networks processing two feature sets extracted from the same videos. The features obtained from the convolutions of both networks are subsequently flattened and merged into a single vector, serving as input for the fully connected layers in the binary classification process. This approach demonstrates the potential for effective ASD detection in children through the combination of gait analysis and deep learning techniques
All-optical and logic gate based on semiconductor optical amplifiers for implementing deep recurrent neural networks
No full textThe development of optical logic gates is a key
factor for enabling next generation of computations in the
context of Deep Learning and Quantum Computing. In this
work, we introduce a scheme for the implementation of an alloptical
AND logic gate, which makes use of semiconductor
optical amplifiers (SOA) in cross-phase modulation
configuration combined with an all-optical XOR gate. Our
analysis includes a realistic model of SOA, which considers
also the phase and the delay of the signals. We prove that our
scheme allows us to obtain almost ideal transitions in 3 out of
the 4 situations in a 2-bit logic, with any SOA. The remaining
combination shows a reduction of extinction ratio, which can
still be improved with better tuning of SOAs
A nonuniform quantizer for hardware implementation of neural networks
No full textNew trends in neural computation, now dealing with distributed learning on pervasive sensor networks and multiple sources of big data, make necessary the use of computationally efficient techniques to be implemented on simple and cheap hardware architectures. In this paper, a nonuniform quantization at the input layer of neural networks is introduced, in order to optimize their implementation on hardware architectures based on a finite precision arithmetic. Namely, we propose a nonlinear A/D conversion of input signals by considering the actual structure of data to be processed. Random Vector Functional-Link is considered as the reference model for neural networks and a genetic optimization is adopted for determining the quantization levels to be found. The proposed approach is assessed by several experimental results obtained on well-known benchmarks for the general problem of data regression
- …
