Brian McGinley

Maintaining Healthy Population Diversity Using Adaptive Crossover, Mutation, and Selection

This paper presents ACROMUSE, a novel genetic algorithm (GA) which adapts crossover, mutation, and selection parameters. ACROMUSEs objective is to create and maintain a diverse population of highly-fit (healthy) individuals, capable of adapting quickly to fitness landscape change and well-suited to the efficient optimization of multimodal fitness landscapes. A new methodology is introduced for determining standard population diversity (SPD) and an original measure of healthy population diversity (HPD) is proposed. The SPD measure is employed to adapt crossover and mutation, while selection pressure is controlled by adapting tournament size according to HPD. In addition to selection pressure control, ACROMUSE tournament selection selects individuals according to healthy diversity contribution rather than fitness. This proposed selection mechanism simultaneously promotes diversity and fitness within the population. The performance of ACROMUSE is evaluated using various multimodal benchmark functions. Statistically significant results are presented comparing ACROMUSEs fitness and diversity performance to that of several other GAs. By maintaining a diverse population of healthy individuals, ACROMUSE responds to fitness landscape change by restoring better fitness scores faster than other GAs. Analysis of the adaptive operators illustrates that the key benefit of ACROMUSE is the synergy of the operators working together to achieve an effective balance between exploration and exploitation.

Scalable Hierarchical Network-on-Chip Architecture for Spiking Neural Network Hardware Implementations

Spiking neural networks (SNNs) attempt to emulate information processing in the mammalian brain based on massively parallel arrays of neurons that communicate via spike events. SNNs offer the possibility to implement embedded neuromorphic circuits, with high parallelism and low power consumption compared to the traditional von Neumann computer paradigms. Nevertheless, the lack of modularity and poor connectivity shown by traditional neuron interconnect implementations based on shared bus topologies is prohibiting scalable hardware implementations of SNNs. This paper presents a novel hierarchical network-on-chip (H-NoC) architecture for SNN hardware, which aims to address the scalability issue by creating a modular array of clusters of neurons using a hierarchical structure of low and high-level routers. The proposed H-NoC architecture incorporates a spike traffic compression technique to exploit SNN traffic patterns and locality between neurons, thus reducing traffic overhead and improving throughput on the network. In addition, adaptive routing capabilities between clusters balance local and global traffic loads to sustain throughput under bursting activity. Analytical results show the scalability of the proposed H-NoC approach under different scenarios, while simulation and synthesis analysis using 65-nm CMOS technology demonstrate high-throughput, low-cost area, and power consumption per cluster, respectively.

Compressed Sensing for Bioelectric Signals: A Review

This paper provides a comprehensive review of compressed sensing or compressive sampling (CS) in bioelectric signal compression applications. The aim is to provide a detailed analysis of the current trends in CS, focusing on the advantages and disadvantages in compressing different biosignals and its suitability for deployment in embedded hardware. Performance metrics such as percent root-mean-squared difference (PRD), signal-to-noise ratio (SNR), and power consumption are used to objectively quantify the capabilities of CS. Furthermore, CS is compared to state-of-the-art compression algorithms in compressing electrocardiogram (ECG) and electroencephalography (EEG) as examples of typical biosignals. The main technical challenges associated with CS are discussed along with the predicted future trends.

A reconfigurable and biologically inspired paradigm for computation using network-on-chip and spiking neural networks

FPGA devices have emerged as a popular platform for the rapid prototyping of biological Spiking Neural Networks (SNNs) applications, offering the key requirement of reconfigurability. However, FPGAs do not efficiently realise the biologically plausible neuron and synaptic models of SNNs, and current FPGA routing structures cannot accommodate the high levels of interneuron connectivity inherent in complex SNNs. This paper highlights and discusses the current challenges of implementing scalable SNNs on reconfigurable FPGAs. The paper proposes a novel field programmable neural network architecture (EMBRACE), incorporating low-power analogue spiking neurons, interconnected using a Network-on-Chip architecture. Results on the evaluation of the EMBRACE architecture using the XOR benchmark problem are presented, and the performance of the architecture is discussed. The paper also discusses the adaptability of the EMBRACE architecture in supporting fault tolerant computing.

EEG compression using JPEG2000: How much loss is too much?

Compression of biosignals is an important means of conserving power in wireless body area networks and ambulatory monitoring systems. In contrast to lossless compression techniques, lossy compression algorithms can achieve higher compression ratios and hence, higher power savings, at the expense of some degradation of the reconstructed signal. In this paper, a variant of the lossy JPEG2000 algorithm is applied to Electroencephalogram (EEG) data from the Freiburg epilepsy database. By varying compression parameters, a range of reconstructions of varying signal fidelity is produced. Although lossy compression has been applied to EEG data in previous studies, it is unclear what level of signal degradation, if any, would be acceptable to a clinician before diagnostically significant information is lost. In this paper, the reconstructed EEG signals are applied to REACT, a state-of-the-art seizure detection algorithm, in order to determine the effect of lossy compression on its seizure detection ability. By using REACT in place of a clinician, many hundreds of hours of reconstructed EEG data are efficiently analysed, thereby allowing an analysis of the amount of EEG signal distortion that can be tolerated. The corresponding compression ratios that can be achieved are also presented.

Spiking Neural Networks for Breast Cancer Classification in a Dielectrically Heterogeneous Breast

The considerable overlap in the dielectric properties of benign and malignant tissue at microwave frequencies means that breast tumour classiflcation using traditional UWB Radar imaging algorithms could be very problematic. Several studies have examined the possibility of using the Radar Target Signature (RTS) of a tumour to classify the tumour as either benign or malignant, since the RTS has been shown to be in∞uenced by the size, shape and surface texture of tumours. The main weakness of existing studies is that they mainly consider tumours in a 3D dielectrically homogenous or 2D heterogeneous breast model. In this paper, the efiects of dielectric heterogeneity on a novel Spiking Neural Network (SNN) classifler are examined in terms of both sensitivity and speciflcity, using a 3D dielectrically heterogeneous breast model. The performance of the SNN classifler is compared to an existing LDA classifler. The efiect of combining con∞icting classiflcation readings in a multi-antenna system is also considered. Finally and importantly, misclassifled tumours are analysed and suggestions for future work are discussed.

Hardware spiking neural network prototyping and application

EMBRACE has been proposed as a scalable, reconfigurable, mixed signal, embedded hardware Spiking Neural Network (SNN) device. EMBRACE, which is yet to be realised, targets the issues of area, power and scalability through the use of a low area, low power analogue neuron/synapse cell, and a digital packet-based Network on Chip (NoC) communication architecture. The paper describes the implementation and testing of EMBRACE-FPGA, an FPGA-based hardware SNN prototype. The operation of the NoC inter-neuron communication approach and its ability to support large scale, reconfigurable, highly interconnected SNNs is illustrated. The paper describes an integrated training and configuration platform and an on-chip fitness function, which supports GA-based evolution of SNN parameters. The practicalities of using the SNN development platform and SNN configuration toolset are described. The paper considers the impact of latency jitter noise introduced by the NoC router and the EMBRACE-FPGA processor-based neuron/synapse model on SNN accuracy and evolution time. Benchmark SNN applications are described and results demonstrate the evolution of high quality and robust solutions in the presence of noise. The reconfigurable EMBRACE architecture enables future investigation of adaptive hardware applications and self repair in evolvable hardware.

The Effects of Lossy Compression on Diagnostically Relevant Seizure Information in EEG Signals

This paper examines the effects of compression on electroencephalogram (EEG) signals, in the context of automated detection of epileptic seizures. Specifically, it examines the use of lossy compression on EEG signals in order to reduce the amount of data which has to be transmitted or stored, while having as little impact as possible on the information in the signal relevant to diagnosing epileptic seizures. Two popular compression methods, JPEG2000 and SPIHT, were used. A range of compression levels was selected for both algorithms in order to compress the signals with varying degrees of loss. This compression was applied to the database of epileptiform data provided by the University of Freiburg, Germany. The real-time EEG analysis for event detection automated seizure detection system was used in place of a trained clinician for scoring the reconstructed data. Results demonstrate that compression by a factor of up to 120:1 can be achieved, with minimal loss in seizure detection performance as measured by the area under the receiver operating characteristic curve of the seizure detection system.

An evaluation of the effects of wavelet coefficient quantisation in transform based EEG compression

In recent years, there has been a growing interest in the compression of electroencephalographic (EEG) signals for telemedical and ambulatory EEG applications. Data compression is an important factor in these applications as a means of reducing the amount of data required for transmission. Allowing for a carefully controlled level of loss in the compression method can provide significant gains in data compression. Quantisation is easy to implement method of data reduction that requires little power expenditure. However, it is a relatively simple, non-invertible operation, and reducing the bit-level too far can result in the loss of too much information to reproduce the original signal to an appropriate fidelity. Other lossy compression methods allow for finer control over compression parameters, generally relying on discarding signal components the coder deems insignificant. SPIHT is a state of the art signal compression method based on the Discrete Wavelet Transform (DWT), originally designed for images but highly regarded as a general means of data compression. This paper compares the approaches of compression by changing the quantisation level of the DWT coefficients in SPIHT, with the standard thresholding method used in SPIHT, to evaluate the effects of each on EEG signals. The combination of increasing quantisation and the use of SPIHT as an entropy encoder has been shown to provide significantly improved results over using the standard SPIHT algorithm alone.

Investigating the Suitability of FPAAs for Evolved Hardware Spiking Neural Networks

This paper investigates the use of a network of cascaded Field Programmable Analogue Arrays (FPAAs) to implement an evolved, analogue, Spiking Neural Network (SNN) pole balance controller. The SNN hardware platform interfaces to a simulated pole balancing model for evaluation. Performance of the evolved analogue hardware controller is compared to that of a software-based SNN controller. The evolved hardware network displays an improved tolerance to changing environments compared with networks evolved solely in simulation. The paper goes on to discuss the suitability of low density FPAA devices for analogue-centric hardware neural network platforms. It concludes by outlining some possible directions which address the observed limitations of using FPAAs for ANNs.