Pasar al contenido principal

SISTEMAS NEUROMORFICOS PARA VISION ARTIFICIAL

PID2019-109465RB-I00

•
Nombre agencia financiadora Agencia Estatal de Investigación
Acrónimo agencia financiadora AEI
Programa Programa Estatal de I+D+i Orientado a los Retos de la Sociedad
Convocatoria Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Año convocatoria 2019
Plan Estatal 2017-2020 Plan Estatal de Investigación Científica y Técnica y de Innovación
Centro beneficiario UNIVERSIDAD DE CADIZ

Resultados relacionados

Resultados totales (Incluyendo duplicados): 11
Encontrada(s) 1 página(s)

RODIN. Repositorio de Objetos de Docencia e Investigación de la Universidad de Cádiz
Publicación Artículo científico (JournalArticle)

WiN-GUI: A graphical tool for neuron-based encoding

  • Müller-Cleve, Simon F.
  • Quintana Velázquez, Fernando Manuel
  • Fra, Vittorio
  • Galindo Riaño, Pedro Luis
  • Pérez Peña, Fernando
  • Urgese, Gianvito
  • Bartolozzi, Chiara
Neuromorphic computing relies on event-based, energy-efficient communication, inherently implying the need for conversion between real-valued (sensory) data and binary, sparse spiking representation. This is usually accomplished by using the real-valued data as current input to a spiking neuron model and tuning the neuron's parameters to match a desired – often biologically inspired – behavior. To support the investigation of neuron models and parameter combinations to identify suitable configurations for neuron-based encoding of sample-based data into spike trains we developed the WiN-GUI. Thanks to the generalized LIF model implemented by default, next to the LIF and Izhikevich neuron models, many spiking behaviors can be investigated out of the box offering the possibility of tuning biologically plausible responses to the input data. The GUI is provided open source and with documentation and is easy to extend with further neuron models and personalize with data analysis functions.
RODIN. Repositorio de Objetos de Docencia e Investigación de la Universidad de Cádiz
Publicación Contribución de congreso (ConferenceOutput)

Event-Based Regression with Spiking Networks

  • Guerrero Vázquez, Elisa
  • Quintana Velázquez, Fernando Manuel
  • Guerrero Lebrero, María de la Paz
Spiking Neuron Networks (SNNs), also known as the third generation of neural networks, are inspired from natural computing in the brain and recent advances in neuroscience. SNNs can overcome the computational power of neural networks made of threshold or sigmoidal units. Recent advances on event-based devices along with their great power, considering the time factor, make SNNs a cutting-edge priority research objective. SNNs have been used mainly for classification problems, but their application to regression tasks remains challenging due to the complexity of training with continuous output data. In the literature we can find some first approximations in regression, specifically, for problems of a single variable of continuous values. This work deals with the analysis of the behavior of SNNs as predictors of multivariable continuous values. For this, a data set based on events has been generated from a bouncing ball and an event-based camera. The goal is to predict the next position of the ball over time.
RODIN. Repositorio de Objetos de Docencia e Investigación de la Universidad de Cádiz
Publicación Artículo científico (JournalArticle)

NESIM-RT: A real-time distributed spiking neural network simulator

  • Rosa-Gallardo, Daniel J.
  • Torre Macías, Juan Carlos de la
  • Quintana Velázquez, Fernando Manuel
  • Domínguez Morales, Juan P.
  • Pérez Peña, Fernando
The neuromorphic engineering field aims to mimic complex biological structures by means of mathematical models, implementing them in either analog or digital circuits. These models include the dynamic characteristics of biological neurons and synapses, which are interconnected creating spiking neural networks. These are first simulated using software frameworks in order to verify the expected behavior of the implemented model. In this work, a novel software, called NESIM-RT, for simulating spiking neural networks in real time is presented. A friendly user interface allows users to design and modify the network, together with visualising its output in real time. A novel implementation of the MDHCP protocol is proposed to support online changes in the network together with allowing the simulation in a many-cores distributed architecture. The performance of the proposed software is compared with other widely-used simulators in the neuromorphic engineering field, highlighting the advantages in terms of latencies and network scalability. NESIM-RT also allows exporting the SNN model directly to SpiNNaker, enabling an immediate transition between software simulation and hardware implementation, Fernando M. Quintana and Juan Carlos de la Torre are funded by the Spanish Ministerio de Ciencia, Innovación Universidades under the FPU grants FPU18/04321 and FPU17/00563 . This work was partially supported by the Spanish grants (with support from the European Regional Development Fund) NEMOVISION ( PID2019-109465RB-I00 ), OPTIMALE ( FEDER-UCA18-108393 ) and MIND-ROB ( PID2019-105556GB-C33 ), together with the Andalusian Regional Project PAIDI2020 GENIUS ( P18-FR-2399 ). Besides, this publication is part of the project TED2021-131880B-I00 supported by the Spanish Ministry (with support from the European Union “NextGenerationEU”). All the authors want to thank Juan Andres Herrera Rodriguez for his support to the manuscript.
RODIN. Repositorio de Objetos de Docencia e Investigación de la Universidad de Cádiz
Publicación Artículo científico (JournalArticle)

Optimal battery management strategies for plug-in electric hybrid buses on routes including green corridors

  • Ruiz Villalobos, Patricia
  • Aragón Jurado, José Miguel
  • Seredynski, M.
  • Cabrera Sánchez, Juan Francisco
  • Peña Morales, David
  • Torre Macías, Juan Carlos de la
  • Zomaya, Albert Y.
  • Dorronsoro Díaz, Bernabé
Public transport is a cornerstone in the transition towards sustainable cities. Moreover, greenhouse gas emissions can be further reduced through powertrain electrification. In this context, plug-in electric hybrid buses emerge as a suitable and flexible solution. They can switch between an electric motor and a combustion engine during operation. An optimal electric drive assignment strategy allows achieving a high electric range and reduced tailpipe emissions. In this work, we look for optimal strategies for maximizing the distance traversed in electric mode and minimizing the total emissions, for real routes including green corridors where the combustion engine cannot be used. Contrary to existing works, this approach does not only focus on the improvement of the bus performance in terms of energy consumption, but also on the environmental benefits and livability of cities. This challenge is solved using two multi-objective state-of-the-art evolutionary algorithms, and a novel heuristic, GreenK. Two real- world scenarios are analyzed, namely bus routes M6 in Badalona, and 18 in Grudziadz. Results show a significant reduction in emissions of up to 21% with respect to the strategy found by GreenK, meaning 24 kg less pollutants emitted daily and over 22.5% electric range increase, compared to the currently deployed solution
RODIN. Repositorio de Objetos de Docencia e Investigación de la Universidad de Cádiz
Publicación Artículo científico (JournalArticle)

Automatic Software Tailoring for Optimal Performance

  • Aragón Jurado, José Miguel
  • Torre Macías, Juan Carlos de la
  • Ruiz Villalobos, Patricia
  • Galindo Riaño, Pedro Luis
  • Zomaya, Albert Y.
  • Dorronsoro Díaz, Bernabé
Efficient green software solutions require being aware of the characteristics of both the software and the hardware where it is executed. Separately optimizing them leads to inefficient results, and there is a need for a perfect synergy between software and hardware for optimal outcomes. We present a novel combinatorial optimization problem for the minimization of the software execution time on a specific hardware, taking into account the existing uncertainty in the system. A solution to the problem is a sequence of LLVM code transformations, and a cellular genetic algorithm is used to find it. Assuming that hardware does not change, reducing the software runtime typically leads to a greener version with lower consumption. To cope with the uncertainty, two novel approaches relying on bootstrap method to compute confident intervals of the software runtime at negligible cost are proposed and compared to three other techniques and -O3 Clang compilation flag over four hardware architectures. Results show how the proposed approach effectively copes with the uncertainty, providing more robust solutions with respect to the compared methods. The execution time of the raw program is reduced from 28.1% to up to 63.2%, outperforming -O3 flag by 13.9% to 26.3%, for the different architectures., This work was supported by the Junta de AndalucÍa and ERDF (GENIUS – P18-2399), the Spanish Ministerio de Ciencia, Innovación y Universidades and ERDF (NEMOVISION– PID2019-109465RB-I00) and ERDF (OPTIMALE – FEDERUCA18-108393). This publication is part of TED2021-131880B-I00 project (eFracWare), funded MCIN/AEI/10. 13039/501100011033 and the European Union “NextGenerationEU”/PRTR and project eMob (PID2022-137858OBI00) funded by Spanish MCIN, the AEI and the ERDF on MCIN/AEI/10.13039/501100011033/FEDER, UE. This work was partially supported by the Spanish Ministry of Science and Innovation and ERDF (RED2018-102472-T). J.C. de la Torre and J.M. Aragón-Jurado would like to acknowledge the Spanish Ministerio de Ciencia, Innovación y Universidades for the support through FPU17/00563 and FPU21/02026 grants, respectively. J.M. Aragón-Jurado acknowledges the grant from Plan Propio - UCA 2022-2023. B. Dorronsoro and P. Ruiz acknowledge “ayuda de recualificación” funded by Ministerio de Universidades and the European Union-NextGenerationEU
RODIN. Repositorio de Objetos de Docencia e Investigación de la Universidad de Cádiz
Publicación Artículo científico (JournalArticle)

Multi-Signal Multifractal Detrended Fluctuation Analysis for Uncertain Systems —Application to the Energy Consumption of Software Programs in Microcontrollers

  • Torre Macías, Juan Carlos de la
  • Pavón Domínguez, Pablo
  • Dorronsoro Díaz, Bernabé
  • Galindo Riaño, Pedro Luis
  • Ruiz Villalobos, Patricia
Uncertain systems are those wherein some variability is observed, meaning that different observations of the system will produce different measurements. Studying such systems demands the use of statistical methods over multiple measurements, which allows overcoming the uncertainty, based on the premise that a single measurement is not representative of the system’s behavior. In such cases, the current multifractal detrended fluctuation analysis (MFDFA) method cannot offer confident conclusions. This work presents multi-signal MFDFA (MS-MFDFA), a novel methodology for accurately characterizing uncertain systems using the MFDFA algorithm, which enables overcoming the uncertainty of the system by simultaneously considering a large set of signals. As a case study, we consider the problem of characterizing software (Sw) consumption. The difficulty of the problem mainly comes from the complexity of the interactions between Sw and hardware (Hw), as well as from the high uncertainty level of the consumption measurements, which are affected by concurrent Sw services, the Hw, and external factors such as ambient temperature. We apply MS-MFDFA to generate a signature of the Sw consumption profile, regardless of the execution time, the consumption levels, and uncertainty. Multiple consumption signals (or time series) are built from different Sw runs, obtaining a high frequency sampling of the instant input current for each of them while running the Sw. A benchmark of eight Sw programs for analysis is also proposed. Moreover, a fully functional application to automatically perform MS-MFDFA analysis has been made freely available. The results showed that the proposed methodology is a suitable approximation for the multifractal analysis of a large number of time series obtained from uncertain systems. Moreover, analysis of the multifractal properties showed that this approach was able to differentiate between the eight Sw programs studied, showing differences in the temporal scaling range where multifractal behavior is found., This publication was supported by the projects eFracWare [TED2021-131880B-I00] funded by the Spanish MCIN and the European Union “NextGenerationEU”/PRTR on MCIN/AEI/10.13039/501100011033, and the project eMob [PID2022-137858OB-I00] funded by the Spanish MCIN, and the Agency and the European Regional Development on MCIN/AEI/10.13039/501100011033/FEDER, UE. This work was also supported by Junta de Andalucía and ERDF under contract P18-2399 (GENIUS), ERDF [OPTIMALE – FEDER-UCA18-108393], and the Spanish Ministerio de Ciencia, Innovación y Universidades and the ERDF [NEMOVISION PID2019-109465RB-I00]. J.C. de la Torre would like to acknowledge the Spanish Ministerio de Ciencia, Innovación y Universidades for the support through FPU grant [FPU17/00563]. B. Dorronsoro and P. Ruiz acknowledge “ayuda de recualificación” funding by Ministerio de Universidades and the European Union-NextGenerationEU.
RODIN. Repositorio de Objetos de Docencia e Investigación de la Universidad de Cádiz
Publicación Contribución de congreso (ConferenceOutput)

Deep Spiking Neural Network for object tracking

  • Guerrero Vázquez, Elisa
  • Quintana Velázquez, Fernando Manuel
  • Guerrero Lebrero, María de la Paz
  • Pérez Peña, Fernando
  • Galindo Riaño, Pedro Luis
We are investigating surrogate gradient as optimization methods in Deep SNN for regression prob- lems. A SNN able to detect a ball at high speed is being developed in which the voltage potential of the output neurons correspond, in real time, with its position, making possible its application in robotic systems that require fast object tracking. As a future work, training and validation over the network and dataset design would be performed using PyTorch framework, as well as the deployment of the system into a robotic platform, for object identification and tracking.
RODIN. Repositorio de Objetos de Docencia e Investigación de la Universidad de Cádiz
Publicación Contribución de congreso (ConferenceOutput)

SpikeBALL: Neuromorphic Dataset for Object Tracking

  • Guerrero Vázquez, Elisa
  • Quintana Velázquez, Fernando Manuel
  • Guerrero Lebrero, María de la Paz
Most of widely used datasets are not suitable for Spiking Neural Networks (SNNs) due to the need to encode the static data into spike trains and then put them into the network. In addition, the majority of these datasets have been generated to classify objects and can not be used to solve object tracking problems. Therefore, we propose a new neuromorphic dataset, SpikeBALL, for object tracking that contributes to improve the development of the SNN algorithm for these type of problems.
RODIN. Repositorio de Objetos de Docencia e Investigación de la Universidad de Cádiz
Publicación Artículo científico (JournalArticle)

GAGI: Game engine for Artificial General Intelligence experimentation

  • Torre Macías, Juan Carlos de la
  • Aragón Jurado, José Miguel
  • Crespo Álvarez, Abdón
  • Barcena González, Guillermo
Video games have been in the focus of the research and academic community for the last few years, with the study and experimentation of Artificial General Intelligence (AGI) standing out. AGI experimentation platforms allow to analyze and study, in a visual way, the behavior of different AI agents previously defined. In this work a novel game engine, called GAGI, capable of serving as an AGI experimentation platform is presented. As a game engine, GAGI is able to design and create novel 2D and 3D video games using C++ programming language. Moreover, GAGI provides the user with a unique environment for simulating and studying AI agents inside the created game. Users can deploy multiple AI agents while interacting with them in real time, improving the understanding of their interactions and behaviors. The features of the proposed software is compared against others widely-used game engines in the video games industry as well as in the research community, highlighting the advantages in terms of design capability and AI support. GAGI also offers the possibility to reproduce the experiments, opening up multiple possibilities for the research community.
RODIN. Repositorio de Objetos de Docencia e Investigación de la Universidad de Cádiz
Publicación Artículo científico (JournalArticle)

Release 2.0—NESIM-RT: A real-time distributed spiking neural network simulator

  • Quintana Velázquez, Fernando Manuel
  • Torre Macías, Juan Carlos de la
  • Barcena González, Guillermo
  • Guerrero Lebrero, María de la Paz
  • Guerrero Vázquez, Elisa
NESIM-RT is a specialized tool designed for simulating neuromorphic systems. In this new release we extend its capabilities to include state-of-the art models like the AdexLIF and Izhikevich, and to incorporate dynamic synaptic mechanisms such as Spike-Timing Dependent Plasticity (STDP). With these new features, researchers can now observe in real-time how different parameters influence these models and learning rules, thereby gaining deeper insights into neuronal function and network dynamics., This work was also supported by the project NEMOVISION from the Ministerio de Ciencia e Innovación, PID2019-109465RB-I00/ AEI/10.13039/501100011033 and by the Junta de Andalucía and ERDF (GENIUS –P18-2399), and ERDF (OPTIMALE – FEDER-UCA18-108393). It is also part of the project TED2021-131880B-I00, funded by MCIN/AEI/10.13039/501100011033 and the European Union ‘‘NextGenerationEU’’/PRTR.
RODIN. Repositorio de Objetos de Docencia e Investigación de la Universidad de Cádiz
Publicación Artículo científico (JournalArticle)

Source code obfuscation with genetic algorithms using LLVM code optimizations

  • Torre Macías, Juan Carlos de la
  • Jareño Dorado, Javier
  • Aragón Jurado, José Miguel
  • Varrette, Sébastien
  • Dorronsoro Díaz, Bernabé
With the advent of the cloud computing model allowing a shared access to massive computing facilities, a surging demand emerges for the protection of the intellectual property tied to the programs executed on these uncontrolled systems. If novel paradigm as confidential computing aims at protecting the data manipulated during the execution, obfuscating techniques (in particular at the source code level) remain a popular solution to conceal the purpose of a program or its logic without altering its functionality, thus preventing reverse-engineering on the program even with the help of computing resources. The many advantages of code obfuscation, together with its low cost, makes it a popular technique. This paper proposes a novel methodology for source code obfuscation that can be used together with other traditional obfuscation techniques, making the code more robust against reverse engineering attacks. Three program complexity metrics are used to define three different single-objective combinatorial optimization versions of the problem, which are solved and analysed. Additionally, three multi-objective problems are defined, those considering each of the selected metrics together with the program execution time, in order to avoid strong obfuscations penalizing the performance. The goal of the defined problems is to find sequences of LLVM optimizations that lead to highly obfuscated versions of the original code. These transformations are applied to the back-end pseudo-assembly code (i.e., LLVM Intermediate Representation), thus avoiding any further optimizations by the compiler. Classical genetic algorithms (GAs) are used to solve the studied problems, namely a basic cellular GA for the single-objective problems and the popular NSGA-II for the multi-objective ones. The promising results show the potential of the proposed technique.