Echo State Networks are recurrent neural networks that leverage a random reservoir’s dynamics, training only a simple readout layer. This approach is computationally efficient but limits network performance. By introducing state or output-feedback connections, the reservoir dynamics can be shaped to enhance the network’s ability to capture complex temporal dependencies. Here we propose AFRICO (Adaptive Feedback, Readout and Input with Connectivity Optimisation), a novel training framework for Echo State Networks that adapts input and state-feedback weights via an Extended Kalman Filter, followed by optimisation of a sparse readout layer that selectively connects reservoir states to the output. The key novelty lies in jointly adapting input and state-feedback pathways to shape reservoir dynamics, while separately constructing a sparse task-specific readout. This approach enables the Echo State Network to capture both the internal dynamics and output mapping of the target system. We evaluate AFRICO on synthetic linear and nonlinear systems, as well as in vivo electrophysiological recordings from fly photoreceptors, demonstrating its versatility for general-purpose input–output time-series modelling across both engineering and biological domains. Across varied hyperparameter initialisations, AFRICO achieves up to 88% reduction in Normalised Mean Squared Error compared to Echo State Networks with fixed output-feedback, while maintaining modest computational effort relative to fully trained Recurrent Neural Networks.

Heterogeneity of gene expression is a characteristic of stem cell populations. Here, we review our own findings with human pluripotent stem cells, emphasizing their capacity to occupy distinct metastable substates that influence their differentiation outcomes. Experimental studies, such as single-cell transcriptomics and analyses of marker dynamics, demonstrate the transient and dynamic nature of these substates. Based on a novel approach to model sequences of fluorescent marker distributions over time, we suggest that transitions between these substates may be governed by deterministic chaos in addition to stochastic fluctuations of gene expression, both of which likely contribute to the observed dynamic behavior.

This paper presents a novel approach to Echo State Networks (ESNs) by integrating state-feedback with switching systems theory and the computational efficiency of Reservoir Computing architectures. We introduce an innovative architecture for ESNs, featuring a switching input weight and state-feedback gain, applicable to both linear and non-linear reservoirs. The stability of this architecture is rigorously analysed using the standard Lyapunov function construction method. Additionally, we explore the dynamical properties of a state-feedback ESN architecture used in the literature. The advantages of the proposed switching ESN is demonstrated using electrophysiological recordings in fruitfly photoreceptors.

Macrophages are central innate immune cells whose function declines with age. The molecular mechanisms underlying age -related changes remain poorly understood, particularly in human macrophages. We report a substantial reduction in phagocytosis, migration, and chemotaxis in human monocyte-derived macrophages (MDMs) from older (>50 years old) compared with younger (18-30 years old) donors, alongside downregulation of transcription factors MYC and USF1. In MDMs from young donors, knockdown of MYC or USF1 decreases phagocytosis and chemotaxis and alters the expression of associated genes, alongside adhesion and extracellular matrix remodeling. A concordant dysregulation of MYC and USF1 target genes is also seen in MDMs from older donors. Furthermore, older age and loss of either MYC or USF1 in MDMs leads to an increased cell size, altered morphology, and reduced actin content. Together, these results define MYC and USF1 as key drivers of MDM age -related functional decline and identify downstream targets to improve macrophage function in aging.

Digital filtering is a fundamental technique in digital signal processing, which operates on a digital sequence without any information on how the sequence was generated. This paper proposes a methodology for designing the equivalent of digital filtering for neuromorphic samples, which are a low-power alternative to conventional digital samples. In the literature, filtering using neuromorphic samples is performed by filtering the reconstructed analog signal, which is required to belong to a predefined input space. We show that this requirement is not necessary, and introduce a new method for computing the neuromorphic samples of the filter output directly from the input samples, backed by theoretical guarantees. We show numerically that we can achieve a similar accuracy compared to that of the conventional method. However, given that we bypass the analog signal reconstruction step, our results show significantly reduced computation time for the proposed method and good performance even when signal recovery is not possible.

Although kernel approximation methods have been widely applied to mitigate the O(n(3)) cost of the n x n kernel matrix inverse in Gaussian process methods, they still face computational challenges. The 'residual' matrix between the covariance matrix and the approximating component is often discarded as it prevents the computational cost reduction. In this paper, we propose a computationally efficient Gaussian process approach that achieves better computational efficiency, O(mn(2)), compared with standard Gaussian process methods, when using m << n data. The proposed approach incorporates the 'residual' matrix in its symmetric diagonally dominant form which can be further approximated by the Neumann series. We have validated and compared the approach with full Gaussian process approaches and kernel approximation based Gaussian process variants, both on synthetic and real air quality data. (C) 2021 Elsevier B.V. All rights reserved.

Small-scale spatial variability in NO2 concentrations is analysed with the help of pollution maps. Maps of NO2 estimated by the Airviro dispersion model and land use regression (LUR) model are fused with measured NO2 concentrations from low-cost sensors (LCS), reference sensors and diffusion tubes. In this study, geostatistical universal kriging was employed for fusing (integrating) model estimations with measured NO2 concentrations. The results showed that the data fusion approach was capable of estimating realistic NO2 concentration maps that inherited spatial patterns of the pollutant from the model estimations and adjusted the modelled values using the measured concentrations. Maps produced by the fusion of NO2-LCS with NO2-LUR produced better results, with r-value 0.96 and RMSE 9.09. Data fusion adds value to both measured and estimated concentrations: the measured data are improved by predicting spatiotemporal gaps, whereas the modelled data are improved by constraining them with observed data. Hotspots of NO2 were shown in the city centre, eastern parts of the city towards the motorway (M1) and on some major roads. Air quality standards were exceeded at several locations in Sheffield, where annual mean NO2 levels were higher than 40 µg/m3. Road traffic was considered to be the dominant emission source of NO2 in Sheffield.

Two air pollutants, oxides of nitrogen (NOx) and particulate matter (PM10), are monitored and modelled employing Airviro air quality dispersion modelling system in Sheffield, United Kingdom. The aim is to determine the most significant emission sources and their spatial variability. NOx emissions (ton/year) from road traffic, point and area sources for the year 2017 were 5370, 6774, and 2425, whereas those of PM10 (ton/year) were 345, 1449, and 281, respectively, which are part of the emission database. The results showed three hotspots of NOx, namely the Sheffield City Centre, Darnall and Tinsley Roundabout (M1 J34S). High PM10 concentrations were shown mainly between Sheffield Forgemasters International (a heavy engineering steel company) and Meadowhall Shopping Centre. Several emission scenarios were tested, which showed that NOx concentrations were mainly controlled by road traffic, whereas PM10 concentrations were controlled by point sources. Spatiotemporal variability and public exposure to air pollution were analysed. NOx concentration was greater than 52 µg/m3 in about 8 km2 area, where more than 66 thousand people lived. Models validated by observations can be used to fill in spatiotemporal gaps in measured data. The approach used presents spatiotemporal situation awareness maps that could be used for decision making and improving the urban infrastructure.

Land Use Regression (LUR) based on multiple linear regression model is one of the techniques used most frequently for modelling the spatial variability of air pollution and assessing exposure in urban areas. In this paper, a nonlinear generalised additive model is proposed for LUR and its performance is compared to a linear model in Sheffield, UK for the year 2019. Pollution models were estimated using NO2 measurements obtained from 188 diffusion tubes and 40 low-cost sensors. Performance of the models was assessed by calculating several statistical metrics including correlation coefficient (R) and root mean square error (RMSE). High resolution (100 m × 100 m) maps demonstrated higher levels of NO2 in the city centre, eastern side of the city and on major roads. The results showed that the nonlinear model outperformed the linear counterpart and that the model estimated using NO2 data from diffusion tubes outperformed the models using data from low-cost sensors or both low-cost sensors and diffusion tubes. The proposed method provides a basis for further application of advanced nonlinear modelling approaches to constructing LUR models in urban areas which enable quantifying small scale variability in pollution levels.

Stationary density functions statistically characterize the stabilized behavior of dynamical systems. Instead of temporal sequences of data, stationary densities are observed to determine the unknown transformations, which is called the inverse Frobenius-Perron problem. This paper proposes a new approach to determining the unique map from stationary densities generated by a one-dimensional discrete-time dynamical system driven by an external control input, given the input density functions that are linearly independent. A numerical simulation example is used to validate the effectiveness of the developed approach. (C) 2020 Elsevier B.V. All rights reserved.

The fruit fly is a key model organism for studying the activity of interconnected brain circuits. A large scattered global research community of neurobiologists and neurogeneticists, computational and theoretical neuroscientists, and computer scientists and engineers has been developing a vast trove of experimental and modeling data that has yet to be distilled into new knowledge and understanding of the functional logic of the brain. Developing open shared models, modelling tools and data repositories that can be accessed from anywhere in the world is the necessary engine for accelerating our understanding of how the brain works. To that end we developed the Fruit Fly Brain Observatory (FFBO), the next generation open-source platform to support open, collaborative Drosophila neuroscience research. FFBO provides a (i) hub for storing and integrating fruit fly brain research data from multiple data sources worldwide, (ii) unified repository of tools and methods to build, emulate and compare fruit fly brain models in health and disease, and (iii) an open framework for fruit fly brain data processing and model execution. FFBO provides access to application tools for visualizing, configuring, simulating and analyzing computational models of brain circuits of the (i) cell type map, (ii) connectome, (iii) synaptome, and (iv) activity map using intuitive queries in plain English. Tools are provided to extract the function inherent in these structural maps. All applications can be accessed with any modern browser.

High Performance Computing (HPC) demand is on the rise, particularly for large distributed computing. HPC systems have, by design, very heterogeneous architectures, both in computation and in communication bandwidth, resulting in wide variations in the cost of communications between compute units. If large distributed applications are to take full advantage of HPC, the physical communication capabilities must be taken into consideration when allocating workload. Hypergraphs are good at modelling total volume of communication in parallel and distributed applications. To the best of our knowledge, there are no hypergraph partitioning algorithms to date that are architecture-aware. We propose a novel restreaming hypergraph partitioning algorithm (HyperPRAW) that takes advantage of peer to peer physical bandwidth profiling data to improve distributed applications performance in HPC systems. Our results show that not only the quality of the partitions achieved by our algorithm is comparable with state-of-the-art multilevel partitioning, but that the runtime performance in a synthetic benchmark is significantly reduced in 10 hypergraph models tested, with speedup factors of up to 14x.

We study the recovery of missing data from multiple smart grid datasets within a matrix completion framework. The datasets contain the electrical magnitudes required for monitoring and control of the electricity distribution system. Each dataset is described by a low rank matrix. Different datasets are correlated as a result of containing measurements of different physical magnitudes generated by the same distribution system. To assess the validity of matrix completion techniques in the recovery of missing data, we characterize the fundamental limits when two correlated datasets are jointly recovered. We then proceed to evaluate the performance of Singular Value Thresholding (SVT) and Bayesian SVT (BSVT) in this setting. We show that BSVT outperforms SVT by simulating the recovery for different correlated datasets. The performance of BSVT displays the tradeoff behaviour described by the fundamental limit, which suggests that BSVT exploits the correlation between the datasets in an efficient manner.

Air pollution in large urban areas has become a serious issue due to its negative impacts on human health, building materials, biodiversity and urban ecosystems in both developed and less-wealthy nations. In most large urban areas, especially in developed countries air quality monitoring networks (AQMN) have been established that provide air quality (AQ) data for various purposes, e.g., to monitor regulatory compliance and to assess the effectiveness of control strategies. However, the criteria of structuring the network are currently defined by single questions rather than attempting to create a network to serve multiple functions. Here we propose a methodology supported by numerical, conceptual and GIS frameworks for structuring AQMN using social, environmental and economic indicators as a case study in Sheffield, UK. The main factors used for air quality monitoring station (AQMS) selection are population-weighted pollution concentration (PWPC) and weighted spatial variability (WSV) incorporating population density (social indicator), pollution levels and spatial variability of air pollutant concentrations (environmental indicator). Total number of sensors is decided on the basis of budget (economic indicator), whereas the number of sensors deployed in each output area is proportional to WSV. The purpose of AQ monitoring and its role in determining the location of AQMS is analysed. Furthermore, the existing AQMN is analysed and an alternative proposed following a formal procedure. In contrast to traditional networks, which are structured based on a single AQ monitoring approach, the proposed AQMN has several layers of sensors: Reference sensors recommended by EU and DEFRA, low-cost sensors (LCS) (AQMesh and Envirowatch E-MOTEs) and IoT (Internet of Things) sensors. The core aim is to structure an integrated AQMN in urban areas, which will lead to the collection of AQ data with high spatiotemporal resolution. The use of LCS in the proposed network provides a cheaper option for setting up a purpose-designed network for greater spatial coverage, especially in low- and middle-income countries.

High Performance Computing (HPC) demand is on the rise, particularly for large distributed computing. HPC systems have, by design, very heterogeneous architectures, both in computation and in communication bandwidth, resulting in wide variations in the cost of communications between compute units. If large distributed applications are to take full advantage of HPC, the physical communication capabilities must be taken into consideration when allocating workload. Hypergraphs are good at modelling total volume of communication in parallel and distributed applications. To the best of our knowledge, there are no hypergraph partitioning algorithms to date that are architecture-aware. We propose a novel restreaming hypergraph partitioning algorithm (HyperPRAW) that takes advantage of peer to peer physical bandwidth profiling data to improve distributed applications performance in HPC systems. Our results show that not only the quality of the partitions achieved by our algorithm is comparable with state-of-the-art multilevel partitioning, but that the runtime performance in a synthetic benchmark is significantly reduced in 10 hypergraph models tested, with speedup factors of up to 14x.

In the last decade there has been a surge in the number of big science projects interested in achieving a comprehensive understanding of the functions of the brain, using Spiking Neuronal Network (SNN) simulations to aid discovery and experimentation. Such an approach increases the computational demands on SNN simulators: if natural scale brain-size simulations are to be realized, it is necessary to use parallel and distributed models of computing. Communication is recognized as the dominant part of distributed SNN simulations. As the number of computational nodes increases, the proportion of time the simulation spends in useful computing (computational efficiency) is reduced and therefore applies a limit to scalability. This work targets the three phases of communication to improve overall computational efficiency in distributed simulations: implicit synchronization, process handshake and data exchange. We introduce a connectivity-aware allocation of neurons to compute nodes by modeling the SNN as a hypergraph. Partitioning the hypergraph to reduce interprocess communication increases the sparsity of the communication graph. We propose dynamic sparse exchange as an improvement over simple point-to-point exchange on sparse communications. Results show a combined gain when using hypergraph-based allocation and dynamic sparse communication, increasing computational efficiency by up to 40.8 percentage points and reducing simulation time by up to 73%. The findings are applicable to other distributed complex system simulations in which communication is modeled as a graph network.

Traditional real-time air quality monitoring instruments are expensive to install and maintain; therefore, such existing air quality monitoring networks are sparsely deployed and lack the measurement density to develop high-resolution spatiotemporal air pollutant maps. More recently, low-cost sensors have been used to collect high-resolution spatial and temporal air pollution data in real-time. In this paper, for the first time, Envirowatch E-MOTEs are employed for air quality monitoring as a case study in Sheffield. Ten E-MOTEs were deployed for a year (October 2016 to September 2017) monitoring several air pollutants (NO, NO2, CO) and meteorological parameters. Their performance was compared to each other and to a reference instrument installed nearby. E-MOTEs were able to successfully capture the temporal variability such as diurnal, weekly and annual cycles in air pollutant concentrations and demonstrated significant similarity with reference instruments. NO2 concentrations showed very strong positive correlation between various sensors. Mostly, correlation coefficients (r values) were greater than 0.92. CO from different sensors also had r values mostly greater than 0.92; however, NO showed r value less than 0.5. Furthermore, several multiple linear regression models (MLRM) and generalised additive models (GAM) were developed to calibrate the E-MOTE data and reproduce NO and NO2 concentrations measured by the reference instruments. GAMs demonstrated significantly better performance than linear models by capturing the non-linear association between the response and explanatory variables. The best GAM developed for reproducing NO2 concentrations returned values of 0.95, 3.91, 0.81, 0.005 and 0.61 for factor of two (FAC2), root mean square error (RMSE), coefficient of determination (R2), normalised mean biased (NMB) and coefficient of efficiency (COE), respectively. The low-cost sensors offer a more affordable alternative for providing real-time high-resolution spatiotemporal air quality and meteorological parameter data with acceptable performance.

There is extensive evidence that biological neural networks encode information in the precise timing of the spikes generated and transmitted by neurons, which offers several advantages over rate-based codes. Here we adopt a vector space formulation of spike train sequences and introduce a new liquid state machine LSM network architecture and a new forward orthogonal regression algorithm to learn an input-output signal mapping or to decode the brain activity. The proposed algorithm uses precise spike timing to select the presynaptic neurons relevant to each learning task. We show that using precise spike timing to train the LSM and selecting the readout presynaptic neurons leads to a significant increase in performance on binary classification tasks, in decoding neural activity from multielectrode array recordings, as well as in a speech recognition task, compared with what is achieved using the standard architecture and training methods.

The advanced operation of future electricity distribution systems is likely to require significant observability of the different parameters of interest (e.g., demand, voltages, currents, etc.). Ensuring completeness of data is, therefore, paramount. In this context, an algorithm for recovering missing state variable observations in electricity distribution systems is presented. The proposed method exploits the low rank structure of the state variables via a matrix completion approach incorporating prior knowledge in the form of second order statistics. Essentially, the recovery method combines nuclear norm minimization with Bayesian estimation. The performance of the new algorithm is compared to the information-theoretic limits and tested through simulations using real data of an urban low voltage distribution system. The impact of the prior knowledge is analyzed when a mismatched covariance is used and under a Markovian sampling that introduces structure in the observation pattern. Numerical results demonstrate that the proposed algorithm is robust and outperforms existing state of the art algorithms.