Andres Upegui | University Of Appleid Scienecs Of Western Switzerland (original) (raw)

Papers by Andres Upegui

2022 25th Euromicro Conference on Digital System Design (DSD)

Zenodo (CERN European Organization for Nuclear Research), Sep 28, 2022

2019 International Joint Conference on Neural Networks (IJCNN), 2019

During the last years, Deep Neural Networks have reached the highest performances in image classi... more During the last years, Deep Neural Networks have reached the highest performances in image classification. Nevertheless, such a success is mostly based on supervised and off-line learning: they require thus huge labeled datasets for learning, and once it is done, they cannot adapt to any change in the data from the environment. In the context of brain-inspired computing, we apply Kohonen-based Self-Organizing Maps for unsupervised learning without labels, and we explore original extensions such as the Dynamic SOM that enables continuous learning and the Pruning Cellular SOM that includes synaptic pruning in neuromorphic circuits. After presenting the three models and the experimental setup for MNIST classification, we compare different methods for automatic labeling based on very few labeled data (1% of the training dataset), and then we compare the performances of the three Kohonen-based Self-Organizing Maps with STDP-based Spiking Neural Networks in terms of accuracy, dynamicity and scalability. Index Terms-brain-inspired computing, self-organizing maps, unsupervised learning, embedded image classification.

2023 24th International Symposium on Quality Electronic Design (ISQED)

Ontogenetic hardware, along with epigenetic (neural) hardware and phylogenetic (evolvable) hardwa... more Ontogenetic hardware, along with epigenetic (neural) hardware and phylogenetic (evolvable) hardware, are the key representatives of a new hardware conception paradigm known as bio-inspired hardware. Ontogenesis is the process that allows living beings to develop by means of mechanisms as growing, self-replication, and self-repair. During the last few years, such ontogenetic mechanisms have been presented as a solution for the design of complex electronic circuits, with the goal of coping with the increasing complexity envisioned for future nano-technology devices. This paper presents an ontogenetic mechanism that allows a system, implemented in a reconfigurable device, to self-replicate, generating an identical copy of itself, by partially self-reconfiguring the device containing it in a dynamic way.

Foraging has been identified as a benchmark for collective robotics. It consists on exploring an ... more Foraging has been identified as a benchmark for collective robotics. It consists on exploring an area and gathering prespecified objects from the environment. In addition to efficiently exploring an area, foragers have to be able to find special targets which are common to the whole population. This work proposes a method to cooperatively perform this particular task. Instead of using local or global localization strategies which can rely on the infrastructure installed in the environment, the proposed approach takes advantage of the knowledge gathered by the population about the localization of the targets. Robots communicate in an instrinsic way the estimation about how near they are from a target, and these estimations guide the navigation of the whole population when looking for these specific areas. The results comprehend some tests assessing the performance, robustness, and scalability of the approach. The proposed approach efficiently guides the robots towards the prespecified targets while allowing the modulation of their speed.

This paper introduces the UbiManager, a tool for managing the ubichip reconfigurable circuit. The... more This paper introduces the UbiManager, a tool for managing the ubichip reconfigurable circuit. The ubichip is a custom reconfigurable electronic device for implementing circuits featuring bio-inspired mechanisms like growth, learning, and evolution. The ubichip has been developed in the framework of Perplexus, a European project that aims to develop a scalable hardware platform made of bio-inspired custom reconfigurable devices for simulating large-scale complex systems. In this paper, we present the software tool used for designing, simulating, emulating, debugging, configuring, and monitoring the systems to be implemented in the ubichip. This paper also presents the dissemination plans of the UbiManager, that consist in a web platform allowing researchers to access the hardware platform from any remote base station.

Self-adaptive autonomous hardware systems require on-chip heuristics to generate the circuit that... more Self-adaptive autonomous hardware systems require on-chip heuristics to generate the circuit that constitutes the desired solution. In this paper, we present a population-oriented hardware architecture for particle swarm optimization with discrete recombination (PSO-DR), a hardware-friendly particle swarm that has shown to perform better than the standard PSO for certain parameter values and test functions. We present simulation and synthesis results showing the feasibility, performance, and advantages of the proposed architecture.

This paper presents a flexible agent-oriented programming framework that provides native support ... more This paper presents a flexible agent-oriented programming framework that provides native support for bio-inspired mechanisms. This solution, developed within the Perplexus IST European project [IST-034632] aims at providing a means for the specification of applications running on a distributed and pervasive network of mobile nodes. In such applications, the deployed systems may face time-changing environments and bio-inspiration may prove useful bringing self-adaptability to the system. The presented framework features are demonstrated on a proof-of-concept application made of simple robots that autonomously improve their behaviour over time.

The ubichip is a bio-inspired reconfigurable circuit developed in the framework of the european p... more The ubichip is a bio-inspired reconfigurable circuit developed in the framework of the european project Perplexus. The ubichip offers special reconfigurability capabilities as self-replication and dynamic routing. This paper describes how to exploit the dynamic routing capabilities of the ubichip in order to implement plastic neural networks. We present an approach for dynamically generating a network topology, where synapses among neurons can be created or destroyed depending on the input stimuli. We describe their implementation in the ubichip, and we analyse the resulting network topology and the network development. This work constitutes a first step toward plastic neural circuits exhibiting more realistic biological features.

The ubichip is a bio-inspired reconfigurable circuit developed in the framework of the european p... more The ubichip is a bio-inspired reconfigurable circuit developed in the framework of the european project Perplexus. The ubichip offers special reconfigurability capabilities, being the dynamic routing one of them. This paper describes how to exploit the dynamic routing capabilities of the ubichip in order to implement synaptogenetic neural networks. We present two techniques for dynamically generating the network topology, we describe their implementation in the ubichip, and we analyse the resulting topology. This work constitutes a first step toward neural circuits exhibiting more realistic neural plasticity features.

Randomly connecting networks have proven to be universal computing machines. By interconnecting a... more Randomly connecting networks have proven to be universal computing machines. By interconnecting a set of nodes in a random way one can model very complicated non-linear dynamic systems. Although random Boolean networks (RBN) use Boolean functions as their basic component, there are not hardware implementations of such systems. The absence of implementations is mainly due to the arbitrary connectionism exhibited by the network, and connection flexibility use to be very expensive in terms of hardware resources. In this paper we present an on-chip self-reconfigurable approach for providing a flexible connectionism at very low resource cost by partially reconfiguring Virtex II FPGAs

The complexity exhibited by pervasive systems is constantly increasing. Customer electronics devi... more The complexity exhibited by pervasive systems is constantly increasing. Customer electronics devices provide day to day a larger amount of functionalities. A common approach for guaranteeing high performance is to include specialized coprocessor units. However, these systems lack flexibility, since one must define, in advance, the coprocessor functionality. A solution to this problem is to use run-time reconfigurable coprocessors, exploiting the advantages of hardware while keeping a flexible platform.

This paper introduces the ubichip, a custom reconfigurable electronic device capable of implement... more This paper introduces the ubichip, a custom reconfigurable electronic device capable of implementing bioinspired circuits featuring growth, learning, and evolution. The ubichip is developed in the framework of Perplexus, a European project that aims to develop a scalable hardware platform made of bio-inspired custom reconfigurable devices for simulating large-scale complex systems. In this paper, we describe the configurability and architectural mechanisms that will allow the implementation of evolvable and developmental cellular and neural systems in an efficient way. These mechanisms are dynamic routing, selfreconfiguration, and a neural-friendly logic cell's architecture.

Microprocessors and Microsystems, 2005

In this paper we present a platform for evolving spiking neural networks on FPGAs. Embedded intel... more In this paper we present a platform for evolving spiking neural networks on FPGAs. Embedded intelligent applications require both high performance, so as to exhibit real-time behavior, and flexibility, to cope with the adaptivity requirements. While hardware solutions offer performance, and software solutions offer flexibility, reconfigurable computing arises between these two types of solutions providing a tradeoff between flexibility and performance. Our platform is described as a combination of three parts: a hardware substrate, a computing engine, and an adaptation mechanism. We present, also, results about the performance and synthesis of the neural network implementation on an FPGA. q

The ubichip is a reconfigurable digital circuit with special reconfiguration mechanisms, such as ... more The ubichip is a reconfigurable digital circuit with special reconfiguration mechanisms, such as dynamic routing and self-replication, for supporting the implementation of bio-inspired hardware systems. The dynamic routing mechanism allows to create and destroy interconnections between remote units in a distributed fashion, thus proving useful for implementing cellular systems featuring dynamic topologies. Evolutionary graph theory investigates genetic and cultural evolution processes using the mathematical formalism of both evolutionary game and graph theory. Populations are embedded in graphs representing interaction and imitation links. Payoffs are assigned and successful individuals are imitated with high probability. This paper describes the hardware implementation of a general evolutionary graph model in which the imitation network changes over time by exploiting the dynamic routing capabilities of the ubichip. As a particular example, we analyze the case of a coordination game played by agents arranged in a cycle in which imitation links are randomly created so as to simulate dynamic small-world networks.

Evolvable Hardware arises as a promising solution for automatic digital synthesis of digital and ... more Evolvable Hardware arises as a promising solution for automatic digital synthesis of digital and analog circuits. During the last decade, a special interest has been focused on evolving digital systems by directly mapping a chromosome on the FPGA configuration bitstream. This approach allowed a great degree of flexibility for evolving circuits. Nowadays, FPGAs routing scheme does not allow doing it in such flexible and safe way, so additional constraints must be introduced. In this paper we summarize three techniques for performing hardware evolution by exploiting the capacities of Virtex families. Among our proposals there are high and low level approaches, and coarse and fine grained components. A modular based evolution, with pre-placed and routed components, provides a coarse grain approach. Two techniques for directly modifying LUT contents on hard macros provide a fine grained evolution. Finally, integrating both approaches, coarse and fine grain, provides a more general and powerful framework.

Foraging has been identified as a benchmark for collective robotics. It consists on exploring an ... more Foraging has been identified as a benchmark for collective robotics. It consists on exploring an area and gathering prespecified objects from the environment. In addition to efficiently exploring an area, foragers have to be able to find special targets which are common to the whole population. This work proposes a method to cooperatively perform this particular task. Instead of using local or global localization strategies which can rely on the infrastructure installed in the environment, the proposed approach takes advantage of the knowledge gathered by the population about the localization of the targets. Robots communicate in an instrinsic way the estimation about how near they are from a target, and these estimations guide the navigation of the whole population when looking for these specific areas. The results comprehend some tests assessing the performance, robustness, and scalability of the approach. The proposed approach efficiently guides the robots towards the prespecified targets while allowing the modulation of their speed.

There is no systematic way to define the optimal topology of an artificial neural network for a g... more There is no systematic way to define the optimal topology of an artificial neural network for a given task. Heuristic methods, such as genetic algorithms, have been widely used to determine the number of neurons and the connectivity required for specific applications. However, artificial evolution uses to be highly time-consuming, making it unsuitable for on-line execution. Herein we present a methodology to evolve neural topologies on digital hardware systems. Evolution is performed on line thanks to the partial reconfiguration properties of Virtex II FPGAs. The genome encodes the combination of different layers, which, once downloaded to the FPGA, compose a neural network. The genetic algorithm execution time is reduced, since the fitness is computed on hardware and the downloaded configuration streams have a reduced size.

In this paper we introduce the Bio-inspired Agent Framework (BAF) developed within the Perplexus ... more In this paper we introduce the Bio-inspired Agent Framework (BAF) developed within the Perplexus IST European project 1 . This BAF is FIPA (Foundation for Intelligent and Physical Agents) compliant as based on the JADE Multi-Agent Platform, portable and suitable for Adhoc networks of mobile nodes (MANET). Its bio-inspired capabilities and reliability services provide a powerful bioinspired distributed tool that opens interesting perspectives for adaptive sensor networks.

2022 25th Euromicro Conference on Digital System Design (DSD)

Zenodo (CERN European Organization for Nuclear Research), Sep 28, 2022

2019 International Joint Conference on Neural Networks (IJCNN), 2019

During the last years, Deep Neural Networks have reached the highest performances in image classi... more During the last years, Deep Neural Networks have reached the highest performances in image classification. Nevertheless, such a success is mostly based on supervised and off-line learning: they require thus huge labeled datasets for learning, and once it is done, they cannot adapt to any change in the data from the environment. In the context of brain-inspired computing, we apply Kohonen-based Self-Organizing Maps for unsupervised learning without labels, and we explore original extensions such as the Dynamic SOM that enables continuous learning and the Pruning Cellular SOM that includes synaptic pruning in neuromorphic circuits. After presenting the three models and the experimental setup for MNIST classification, we compare different methods for automatic labeling based on very few labeled data (1% of the training dataset), and then we compare the performances of the three Kohonen-based Self-Organizing Maps with STDP-based Spiking Neural Networks in terms of accuracy, dynamicity and scalability. Index Terms-brain-inspired computing, self-organizing maps, unsupervised learning, embedded image classification.

2023 24th International Symposium on Quality Electronic Design (ISQED)

Ontogenetic hardware, along with epigenetic (neural) hardware and phylogenetic (evolvable) hardwa... more Ontogenetic hardware, along with epigenetic (neural) hardware and phylogenetic (evolvable) hardware, are the key representatives of a new hardware conception paradigm known as bio-inspired hardware. Ontogenesis is the process that allows living beings to develop by means of mechanisms as growing, self-replication, and self-repair. During the last few years, such ontogenetic mechanisms have been presented as a solution for the design of complex electronic circuits, with the goal of coping with the increasing complexity envisioned for future nano-technology devices. This paper presents an ontogenetic mechanism that allows a system, implemented in a reconfigurable device, to self-replicate, generating an identical copy of itself, by partially self-reconfiguring the device containing it in a dynamic way.

Foraging has been identified as a benchmark for collective robotics. It consists on exploring an ... more Foraging has been identified as a benchmark for collective robotics. It consists on exploring an area and gathering prespecified objects from the environment. In addition to efficiently exploring an area, foragers have to be able to find special targets which are common to the whole population. This work proposes a method to cooperatively perform this particular task. Instead of using local or global localization strategies which can rely on the infrastructure installed in the environment, the proposed approach takes advantage of the knowledge gathered by the population about the localization of the targets. Robots communicate in an instrinsic way the estimation about how near they are from a target, and these estimations guide the navigation of the whole population when looking for these specific areas. The results comprehend some tests assessing the performance, robustness, and scalability of the approach. The proposed approach efficiently guides the robots towards the prespecified targets while allowing the modulation of their speed.

This paper introduces the UbiManager, a tool for managing the ubichip reconfigurable circuit. The... more This paper introduces the UbiManager, a tool for managing the ubichip reconfigurable circuit. The ubichip is a custom reconfigurable electronic device for implementing circuits featuring bio-inspired mechanisms like growth, learning, and evolution. The ubichip has been developed in the framework of Perplexus, a European project that aims to develop a scalable hardware platform made of bio-inspired custom reconfigurable devices for simulating large-scale complex systems. In this paper, we present the software tool used for designing, simulating, emulating, debugging, configuring, and monitoring the systems to be implemented in the ubichip. This paper also presents the dissemination plans of the UbiManager, that consist in a web platform allowing researchers to access the hardware platform from any remote base station.

Self-adaptive autonomous hardware systems require on-chip heuristics to generate the circuit that... more Self-adaptive autonomous hardware systems require on-chip heuristics to generate the circuit that constitutes the desired solution. In this paper, we present a population-oriented hardware architecture for particle swarm optimization with discrete recombination (PSO-DR), a hardware-friendly particle swarm that has shown to perform better than the standard PSO for certain parameter values and test functions. We present simulation and synthesis results showing the feasibility, performance, and advantages of the proposed architecture.

This paper presents a flexible agent-oriented programming framework that provides native support ... more This paper presents a flexible agent-oriented programming framework that provides native support for bio-inspired mechanisms. This solution, developed within the Perplexus IST European project [IST-034632] aims at providing a means for the specification of applications running on a distributed and pervasive network of mobile nodes. In such applications, the deployed systems may face time-changing environments and bio-inspiration may prove useful bringing self-adaptability to the system. The presented framework features are demonstrated on a proof-of-concept application made of simple robots that autonomously improve their behaviour over time.

The ubichip is a bio-inspired reconfigurable circuit developed in the framework of the european p... more The ubichip is a bio-inspired reconfigurable circuit developed in the framework of the european project Perplexus. The ubichip offers special reconfigurability capabilities as self-replication and dynamic routing. This paper describes how to exploit the dynamic routing capabilities of the ubichip in order to implement plastic neural networks. We present an approach for dynamically generating a network topology, where synapses among neurons can be created or destroyed depending on the input stimuli. We describe their implementation in the ubichip, and we analyse the resulting network topology and the network development. This work constitutes a first step toward plastic neural circuits exhibiting more realistic biological features.

The ubichip is a bio-inspired reconfigurable circuit developed in the framework of the european p... more The ubichip is a bio-inspired reconfigurable circuit developed in the framework of the european project Perplexus. The ubichip offers special reconfigurability capabilities, being the dynamic routing one of them. This paper describes how to exploit the dynamic routing capabilities of the ubichip in order to implement synaptogenetic neural networks. We present two techniques for dynamically generating the network topology, we describe their implementation in the ubichip, and we analyse the resulting topology. This work constitutes a first step toward neural circuits exhibiting more realistic neural plasticity features.

Randomly connecting networks have proven to be universal computing machines. By interconnecting a... more Randomly connecting networks have proven to be universal computing machines. By interconnecting a set of nodes in a random way one can model very complicated non-linear dynamic systems. Although random Boolean networks (RBN) use Boolean functions as their basic component, there are not hardware implementations of such systems. The absence of implementations is mainly due to the arbitrary connectionism exhibited by the network, and connection flexibility use to be very expensive in terms of hardware resources. In this paper we present an on-chip self-reconfigurable approach for providing a flexible connectionism at very low resource cost by partially reconfiguring Virtex II FPGAs

The complexity exhibited by pervasive systems is constantly increasing. Customer electronics devi... more The complexity exhibited by pervasive systems is constantly increasing. Customer electronics devices provide day to day a larger amount of functionalities. A common approach for guaranteeing high performance is to include specialized coprocessor units. However, these systems lack flexibility, since one must define, in advance, the coprocessor functionality. A solution to this problem is to use run-time reconfigurable coprocessors, exploiting the advantages of hardware while keeping a flexible platform.

This paper introduces the ubichip, a custom reconfigurable electronic device capable of implement... more This paper introduces the ubichip, a custom reconfigurable electronic device capable of implementing bioinspired circuits featuring growth, learning, and evolution. The ubichip is developed in the framework of Perplexus, a European project that aims to develop a scalable hardware platform made of bio-inspired custom reconfigurable devices for simulating large-scale complex systems. In this paper, we describe the configurability and architectural mechanisms that will allow the implementation of evolvable and developmental cellular and neural systems in an efficient way. These mechanisms are dynamic routing, selfreconfiguration, and a neural-friendly logic cell's architecture.

Microprocessors and Microsystems, 2005

In this paper we present a platform for evolving spiking neural networks on FPGAs. Embedded intel... more In this paper we present a platform for evolving spiking neural networks on FPGAs. Embedded intelligent applications require both high performance, so as to exhibit real-time behavior, and flexibility, to cope with the adaptivity requirements. While hardware solutions offer performance, and software solutions offer flexibility, reconfigurable computing arises between these two types of solutions providing a tradeoff between flexibility and performance. Our platform is described as a combination of three parts: a hardware substrate, a computing engine, and an adaptation mechanism. We present, also, results about the performance and synthesis of the neural network implementation on an FPGA. q

The ubichip is a reconfigurable digital circuit with special reconfiguration mechanisms, such as ... more The ubichip is a reconfigurable digital circuit with special reconfiguration mechanisms, such as dynamic routing and self-replication, for supporting the implementation of bio-inspired hardware systems. The dynamic routing mechanism allows to create and destroy interconnections between remote units in a distributed fashion, thus proving useful for implementing cellular systems featuring dynamic topologies. Evolutionary graph theory investigates genetic and cultural evolution processes using the mathematical formalism of both evolutionary game and graph theory. Populations are embedded in graphs representing interaction and imitation links. Payoffs are assigned and successful individuals are imitated with high probability. This paper describes the hardware implementation of a general evolutionary graph model in which the imitation network changes over time by exploiting the dynamic routing capabilities of the ubichip. As a particular example, we analyze the case of a coordination game played by agents arranged in a cycle in which imitation links are randomly created so as to simulate dynamic small-world networks.

Evolvable Hardware arises as a promising solution for automatic digital synthesis of digital and ... more Evolvable Hardware arises as a promising solution for automatic digital synthesis of digital and analog circuits. During the last decade, a special interest has been focused on evolving digital systems by directly mapping a chromosome on the FPGA configuration bitstream. This approach allowed a great degree of flexibility for evolving circuits. Nowadays, FPGAs routing scheme does not allow doing it in such flexible and safe way, so additional constraints must be introduced. In this paper we summarize three techniques for performing hardware evolution by exploiting the capacities of Virtex families. Among our proposals there are high and low level approaches, and coarse and fine grained components. A modular based evolution, with pre-placed and routed components, provides a coarse grain approach. Two techniques for directly modifying LUT contents on hard macros provide a fine grained evolution. Finally, integrating both approaches, coarse and fine grain, provides a more general and powerful framework.

Foraging has been identified as a benchmark for collective robotics. It consists on exploring an ... more Foraging has been identified as a benchmark for collective robotics. It consists on exploring an area and gathering prespecified objects from the environment. In addition to efficiently exploring an area, foragers have to be able to find special targets which are common to the whole population. This work proposes a method to cooperatively perform this particular task. Instead of using local or global localization strategies which can rely on the infrastructure installed in the environment, the proposed approach takes advantage of the knowledge gathered by the population about the localization of the targets. Robots communicate in an instrinsic way the estimation about how near they are from a target, and these estimations guide the navigation of the whole population when looking for these specific areas. The results comprehend some tests assessing the performance, robustness, and scalability of the approach. The proposed approach efficiently guides the robots towards the prespecified targets while allowing the modulation of their speed.

There is no systematic way to define the optimal topology of an artificial neural network for a g... more There is no systematic way to define the optimal topology of an artificial neural network for a given task. Heuristic methods, such as genetic algorithms, have been widely used to determine the number of neurons and the connectivity required for specific applications. However, artificial evolution uses to be highly time-consuming, making it unsuitable for on-line execution. Herein we present a methodology to evolve neural topologies on digital hardware systems. Evolution is performed on line thanks to the partial reconfiguration properties of Virtex II FPGAs. The genome encodes the combination of different layers, which, once downloaded to the FPGA, compose a neural network. The genetic algorithm execution time is reduced, since the fitness is computed on hardware and the downloaded configuration streams have a reduced size.

In this paper we introduce the Bio-inspired Agent Framework (BAF) developed within the Perplexus ... more In this paper we introduce the Bio-inspired Agent Framework (BAF) developed within the Perplexus IST European project 1 . This BAF is FIPA (Foundation for Intelligent and Physical Agents) compliant as based on the JADE Multi-Agent Platform, portable and suitable for Adhoc networks of mobile nodes (MANET). Its bio-inspired capabilities and reliability services provide a powerful bioinspired distributed tool that opens interesting perspectives for adaptive sensor networks.