No description available.

<div>Computing with Artificial Neural Networks (ANNs) is a branch of machine learning that has seen substantial growth over the last decade, significantly increasing the accuracy and capability of machine learning systems. ANNs are connected networks of computing elements inspired by the neuronal connectivity in the brain. Spiking Neural Networks (SNNs) are a type of ANN that operate with event-driven computation, inspired by the “spikes” or firing events of individual neurons in the brain. Neuromorphic computing—the implementation of neural networks in hardware—seeks to improve the energy efficiency of these machine learning systems either by computing directly with device physical primitives, by bypassing the software layer of logical implementations, or by operating with SNN event-driven computation. Such implementations may, however, have added restrictions, including weight-localized learning and hard-wired connections. Further obstacles, such as catastrophic forgetting, the lack of supervised error signals, and storage and energy constraints, are encountered when these systems need to perform autonomous online, real-time learning in an unknown, changing environment. </div><div><br></div><div>Adapting neural network learning algorithms for these constraints can help address these issues. Specifically, corrections to Spike Timing-Dependent Plasticity (STDP) can stabilize local, unsupervised learning; accounting for the statistical firing properties of spiking neurons may improve conversions from non-spiking to spiking networks; biologically-inspired dopaminergic and habituation adjustments to STDP can limit catastrophic forgetting; convolving temporally instead of spatially can provide for localized weight sharing with direct synaptic connections; and explicitly training for spiking sparsity can significantly reduce computational energy consumption.</div>

Simultaneous Localization And Mapping (SLAM) is an important topic within the field of roboticsaiming to localize an agent in a unknown or partially known environment while simultaneouslymapping the environment. The ability to perform robust SLAM is especially important inhazardous environments such as natural disasters, firefighting and space exploration wherehuman exploration may be too dangerous or impractical. In recent years, neuromorphiccameras have been made commercially available. This new type of sensor does not outputconventional frames but instead an asynchronous signal of events at a microsecond resolutionand is capable of capturing details in complex lightning scenarios where a standard camerawould be either under- or overexposed, making neuromorphic cameras a promising solution insituations where standard cameras struggle. This thesis explores a set of different approachesto virtual frames, a frame-based representation of events, in the context of SLAM.UltimateSLAM, a project fusing events, gray scale and IMU data, is investigated using virtualframes of fixed and varying frame rate both with and without motion compensation. Theresulting trajectories are compared to the trajectories produced when using gray scale framesand the number of detected and tracked features are compared. We also use a traditional visualSLAM project, ORB-SLAM, to investigate the Gaussian weighted virtual frames and gray scaleframes reconstructed from the event stream using a recurrent network model. While virtualframes can be used for SLAM, the event camera is not a plug and play sensor and requires agood choice of parameters when constructing virtual frames, relying on pre-existing knowledgeof the scene.

It is believed that neuromorphic hardware will accelerate neuroscience research and enable the next generation edge AI. On the other hand, brain-inspired algorithms are supposed to work efficiently on neuromorphic hardware. But both processes don't happen automatically. To efficiently bring together hardware and algorithm, optimizations are necessary based on the understanding of both sides. In this work, software frameworks and optimizations for efficient implementation of neural network-based algorithms on SpiNNaker 2 are proposed, resulting in optimized power consumption, memory footprint and computation time. In particular, first, a software framework including power management strategies is proposed to apply dynamic voltage and frequency scaling (DVFS) to the simulation of spiking neural networks, which is also the first-ever software framework running a neural network on SpiNNaker 2. The result shows the power consumption is reduced by 60.7% in the synfire chain benchmark. Second, numerical optimizations and data structure optimizations lead to an efficient implementation of reward-based synaptic sampling, which is one of the most complex plasticity algorithms ever implemented on neuromorphic hardware. The results show a reduction of computation time by a factor of 2 and energy consumption by 62%. Third, software optimizations are proposed which effectively exploit the efficiency of the multiply-accumulate array and the flexibility of the ARM core, which results in, when compared with Loihi, 3 times faster inference speed and 5 times lower energy consumption in a keyword spotting benchmark, and faster inference speed and lower energy consumption for adaptive control benchmark in high dimensional cases. The results of this work demonstrate the potential of SpiNNaker 2, explore its range of applications and also provide feedback for the design of the next generation neuromorphic hardware.

This work presents a step forward in the use of 2D layered materials (2DLM), specifically hexagonal boron nitride (h-BN), for the fabrication of memristors. In this study, we fabricate, characterize, and use h-BN based memristors with Ag/few-layer h-BN/Ag structure to implement a fully functioning artificial leaky integrate-and-fire neuron on hardware. The devices showed volatile resistive switching behavior with no electro-forming process required, with relatively low VSET and long endurance of beyond 1.5 million cycles. In addition, we present some of the failure mechanisms in these devices with some statistical analyses to understand the causes, as well as a statistical study of both cycle-to-cycle and device-to-device variabilities in 20 devices. Moreover, we study the use of these devices in implementing a functioning artificial leaky integrate-and-fire neuron similar to a biological neuron in the brain. We provide SPICE simulation as well as hardware implementation of the artificial neuron that are in full agreement, showing that our device could be used for such application. Additionally, we study the use of these devices as an activation function for spiking neural networks (SNNs) by providing a SPICE simulation of a fully trained network, where the artificial spiking neuron is connected to the output terminal of a crossbar array. The SPICE simulations provide a proof of concept for using h-BN based memristor for activation function for SNNs.

Artificial intelligence is a new paradigm of information processing where machines emulate human intelligence and perform tasks that cannot be done with standard computers. Neuromorphic computing is in particular inspired by how the brain computes. Large network of interconnected neurons whose synapses are varied during a learning phase, and where the information flows in parallel throughout different connections. Photonics platforms represent an interesting possibility where to implement neuromorphic processing schemes, exploiting light and its advantages in terms of speed, low energy consumption and inherent parallelism via wavelength division multiplexing. In particular, a candidate playing a diversity of key roles in integrated networks is the microring resonator. In silicon photonics, the microring resonator can implement the strength of a synapse, the spiking emission of a biological neuron, and it can exhibit a fading memory based on its multiple linear and nonlinear dynamical timescales. This manuscript presents an overview of the main applications of silicon microring resonators in neuromorphic silicon photonics, and then focuses on its implementation in a processing scheme, named time delay reservoir computing (RC). Time delay RC is a hardwarefriendly approach by which implement a large neural network, where this is folded in the nonlinear dynamical response of only one physical node, such as a dynamical system with delay feedback. The manuscript illustrates, both numerically and experimentally, how to make time delay RC exploiting the linear and nonlinear dynamical response of a silicon microring resonator. The microring is coupled to an external optical feedback and the results on a diversity of time series prediction tasks and delayed-boolean tasks are presented. Numerically, it is shown that the microring nonlinearities can be exploited to improve the performance on prediction tasks, such as the Santa Fe and Mackey Glass ones. Experimentally, it is shown how the network can be set to solve delayed boolean tasks with error-free operation, at 12 MHz operational speed, together with possible upgrades and alternative implementations that can boost its performances. / La inteligencia artificial es un nuevo paradigma de procesamiento de información en el que las máquinas emulan la inteligencia humana y realizan tareas que no pueden ser realizadas con ordenadores estándar. La computación neuromórfica está particularmente inspirada en cómo el cerebro realiza cálculos. Consiste en una gran red de neuronas interconectadas cuyas sinapsis varían durante una fase de aprendizaje, y donde la información fluye en paralelo a través de diferentes conexiones. Las plataformas fotónicas representan una interesante posibilidad para implementar esquemas de procesamiento neuromórfico, aprovechando las ventajas de la luz en términos de velocidad, bajo consumo de energía e inherente paralelismo a través de la multiplexación por división de longitud de onda. En particular, un candidato que desempeña una diversidad de roles clave en redes integradas es el micro-anillo resonador. En la fotónica de silicio, el micro-anillo resonador puede implementar la intensidad sináptica, la emisión de pulsos de una neurona biológica, y puede exhibir una memoria que decae con el tiempo basada en sus múltiples escalas temporales dinámicas lineales y no lineales. Esta tesis presenta una visión general de las principales aplicaciones de los resonadores de anillo microscópicos de silicio en la fotónica neuromórfica de silicio y se centra en su implementación en un esquema de procesamiento llamado time delay reservoir computing (RC). Time delay RC es un enfoque favorable para el hardware mediante el cual se implementa una gran red neural, a través de la respuesta dinámica no lineal de solo un nodo físico, como un sistema dinámico sujeto a retroalimentación. Este trabajo ilustra, tanto numérica como experimentalmente, cómo realizar la computación en time delay RC utilizando la respuesta dinámica lineal y no lineal de un resonador de anillo microscópico de silicio. El microanillo resonador está acoplado a una retroalimentación óptica externa y se presentan los resultados de una diversidad de tareas de predicción de series temporales y tareas booleanas retrasadas. Numéricamente, se muestra que las no-linealidades del micro-anillo resonador se pueden aprovechar para mejorar el rendimiento en tareas de predicción, como las de Santa Fe y Mackey Glass. Experimentalmente, se muestra cómo la red se puede configurar para resolver tareas booleanas retrasadas sin errores, a una velocidad operativa de 12 MHz, junto con posibles mejoras e implementaciones alternativas que pueden aumentar su rendimiento.

No description available.

No description available.

No description available.

No description available.