arXiv subjects

Jake Love

Publications and source records attributed to Jake Love.

3 recordsLinked to original sources

arXiv2022

Neuromorphic Overparameterisation and Few-Shot Learning in Multilayer Physical Neural Networks

Physical neuromorphic computing, exploiting the complex dynamics of physical systems, has seen rapid advancements in sophistication and performance. Physical reservoir computing, a subset of neuromorphic computing, faces limitations due to its reliance on single systems. This constrains output dimensionality and dynamic range, limiting performance to a narrow range of tasks. Here, we engineer a suite of nanomagnetic array physical reservoirs and interconnect them in parallel and series to create a multilayer neural network architecture. The output of one reservoir is recorded, scaled and virtually fed as input to the next reservoir. This networked approach increases output dimensionality, internal dynamics and computational performance. We demonstrate that a physical neuromorphic system can achieve an overparameterised state, facilitating meta-learning on small training sets and yielding strong performance across a wide range of tasks. Our approach's efficacy is further demonstrated through few-shot learning, where the system rapidly adapts to new tasks.

cond-mat.mes-hall↗

arXiv2022

Audio Classification with Skyrmion Reservoirs

Physical reservoir computing is a computational paradigm that enables spatio-temporal pattern recognition to be performed directly in matter. The use of physical matter leads the way towards energy-efficient devices capable of solving machine learning problems without having to build a system of millions of interconnected neurons. We propose a high performance "skyrmion mixture reservoir" that implements the reservoir computing model with multi-dimensional inputs. We show that our implementation solves spoken digit classification tasks at the nanosecond timescale, with an overall model accuracy of 97.4% and a less that 1% word error rate; the best performance ever reported for in-materio reservoir computers. Due to the quality of the results and the low power properties of magnetic texture reservoirs, we argue that skyrmion fabrics are a compelling candidate for reservoir computing.

cond-mat.mes-hall↗

arXiv2021

Spatial Analysis of Physical Reservoir Computers

Physical reservoir computing is a computational framework that implements spatiotemporal information processing directly within physical systems. By exciting nonlinear dynamical systems and creating linear models from their state, we can create highly energy-efficient devices capable of solving machine learning tasks without building a modular system consisting of millions of neurons interconnected by synapses. To act as an effective reservoir, the chosen dynamical system must have two desirable properties: nonlinearity and memory. We present task agnostic spatial measures to locally measure both of these properties and exemplify them for a specific physical reservoir based upon magnetic skyrmion textures. In contrast to typical reservoir computing metrics, these metrics can be resolved spatially and in parallel from a single input signal, allowing for efficient parameter search to design efficient and high-performance reservoirs. Additionally, we show the natural trade-off between memory capacity and nonlinearity in our reservoir's behaviour, both locally and globally. Finally, by balancing the memory and nonlinearity in a reservoir, we can improve its performance for specific tasks.

cs.LG↗