A5007091764- Advanced Memory and Neural Computing
- Error Correcting Code Techniques
- Neural Networks and Applications
- Neural Networks and Reservoir Computing
- Stochastic Gradient Optimization Techniques
- Ferroelectric and Negative Capacitance Devices
- Wireless Communication Security Techniques
- Quantum Computing Algorithms and Architecture
- Computational Drug Discovery Methods
- Machine Learning and ELM
- Machine Learning in Materials Science
- Cell Image Analysis Techniques
- Neural dynamics and brain function
Universitat de les Illes Balears
2018-2024
Health Research Institute of the Balearic Islands
2022
Edge artificial intelligence (AI) is receiving a tremendous amount of interest from the machine learning community due to ever-increasing popularization Internet Things (IoT). Unfortunately, incorporation AI characteristics edge computing devices presents drawbacks being power and area hungry for typical deep techniques such as convolutional neural networks (CNNs). In this work, we propose power-and-area efficient architecture based on exploitation correlation phenomenon in stochastic (SC)...
The development of power-efficient Machine Learning Hardware is high importance to provide Artificial Intelligence (AI) characteristics those devices operating at the Edge. Unfortunately, state-of-the-art data-driven AI techniques such as deep learning are too costly in terms hardware and energy requirements for Edge Computing (EC) devices. Recently, Cellular Automata (CA) have been proposed a feasible way implement Reservoir (RC) systems which automaton rule fixed training performed using...
This work aimed to enhance a previous neural network hardware implementation based on an efficient combination of Stochastic Computing (SC) and Morphological Neural Networks (MNN). enhancement focused exploiting the natural ease morphological neurons be pruned in order drastically shrink resources increase compactness our network. That is why we extended original hybrid two-layer classify MNIST problem, much more demanding benchmark with about 160,000 trainable parameters. The 92% weights...
Elementary cellular automata (ECA) is a widely studied one-dimensional processing methodology where the successive iteration of automaton may lead to recreation rich pattern dynamic. Recently, have been proposed as feasible way implement Reservoir Computing (RC) systems in which rule fixed and training performed using linear regression. In this work we perform an exhaustive study performance different ECA rules when applied recognition time-independent input signals RC scheme. Once tested,...
Stochastic computing (SC) is a probabilistic-based processing methodology that has emerged as an energy-efficient solution for implementing image and deep learning in hardware. The core of these systems relies on the selection appropriate Random Number Generators (RNGs) to guarantee acceptable accuracy. In this work, we demonstrate classical Linear Feedback Shift Registers (LFSR) can be efficiently used correlation-sensitive circuits if seed followed. For purpose, implement some basic SC...
In this work we propose a new methodology for neural network hardware implementation based on an efficient combination of Stochastic Computing and Morphological Neural Networks (MNN). The main reasons behind the use morphological neurons instead conventional ones are that activation functions not necessary emerges as natural compact way to implement MNN. proposed design has been verified by implementing classical pattern recognition problems such Fisher's IRIS dataset or handwritten digit...
Deploying modern neural networks on resource-constrained edge devices necessitates a series of optimizations to ready them for production. These typically involve pruning, quantization, and fixed-point conversion compress the model size enhance energy efficiency. While these are generally adequate most devices, there exists potential further improving efficiency by leveraging special-purpose hardware unconventional computing paradigms. In this study, we explore stochastic their impact...
Edge Artificial Intelligence or Intelligence, is beginning to receive a tremendous amount of interest. Unfortunately, the incorporation AI characteristics edge computing devices presents drawbacks being power and area hungry for typical machine learning techniques such as Convolutional Neural Networks (CNN). In this work, we propose new power-and-area-efficient architecture implementing (ANNs) in hardware, based on exploitation correlation phenomenon Stochastic Computing (SC) systems. The...
Stochastic computing is an emerging scientific field pushed by the need for developing high-performance artificial intelligence systems in hardware to quickly solve complex data processing problems. This case of virtual screening, a computational task aimed at searching across huge molecular databases new drug leads. In this work, we show classification framework which molecules are described energy-based vector. vector then processed ultra-fast neural network implemented through FPGA using...
A new trans-disciplinary knowledge area, Edge Artificial Intelligence or Intelligence, is beginning to receive a tremendous amount of interest from the machine learning community due ever increasing popularization Internet Things (IoT). Unfortunately, incorporation AI characteristics edge computing devices presents drawbacks being power and area hungry for typical techniques such as Convolutional Neural Networks (CNN). In this work, we propose power-and-area-efficient architecture...
In this work we present an enhancement of a neural network hardware implementation based on efficient combination Stochastic Computing (SC) and Morphological Neural Networks (MNN). The has concentrated extending original tiny two-layer to the more de-manding MNIST benchmark also pruning up 92% weights morphological layer, allowing drastic shrinkage resources power dissipation without barely degrading test accuracy. This new proposal contributes foster promising ultra-low Machine Learning...
Deploying modern neural networks on resource-constrained edge devices requires a series of optimizations to prepare them for production. These typically involve pruning, quantization, and fixed-point conversion compress the model size improve energy efficiency. While these are generally sufficient most devices, there is potential further improving efficiency by leveraging special-purpose hardware unconventional computing paradigms. In this work, we investigate stochastic their impact...
Artificial Neural Networks (ANN) have been popularized in many science and technological areas due to their capacity solve complex pattern matching problems. That is the case of Virtual Screening, a research area that studies how identify those molecular compounds with highest probability present biological activity for therapeutic target. Due vast number small organic thousands targets which such large-scale screening can potentially be carried out, there has an increasing interest...