Hyperdimensional computing with holographic and adaptive encoder

Introduction

Brain-inspired computing has become an emerging field, where a growing number of works focus on developing algorithms that bring machine learning closer to human brains at the functional level. As one of the promising directions, Hyperdimensional Computing (HDC) is centered around the idea of having holographic and high-dimensional representation as the neural activities in our brains. Such representation is the fundamental enabler for the efficiency and robustness of HDC. However, existing HDC-based algorithms suffer from limitations within the encoder. To some extent, they all rely on manually selected encoders, meaning that the resulting representation is never adapted to the tasks at hand.

Methods

In this paper, we propose FLASH, a novel hyperdimensional learning method that incorporates an adaptive and learnable encoder design, aiming at better overall learning performance while maintaining good properties of HDC representation. Current HDC encoders leverage Random Fourier Features (RFF) for kernel correspondence and enable locality-preserving encoding. We propose to learn the encoder matrix distribution via gradient descent and effectively adapt the kernel for a more suitable HDC encoding.

Results

Our experiments on various regression datasets show that tuning the HDC encoder can significantly boost the accuracy, surpassing the current HDC-based algorithm and providing faster inference than other baselines, including RFF-based kernel ridge regression.

Discussion

The results indicate the importance of an adaptive encoder and customized high-dimensional representation in HDC.

EventHD: Robust and efficient hyperdimensional learning with neuromorphic sensor

Brain-inspired computing models have shown great potential to outperform today's deep learning solutions in terms of robustness and energy efficiency. Particularly, Hyper-Dimensional Computing (HDC) has shown promising results in enabling efficient and robust cognitive learning. In this study, we exploit HDC as an alternative computational model that mimics important brain functionalities toward high-efficiency and noise-tolerant neuromorphic computing. We present EventHD, an end-to-end learning framework based on HDC for robust, efficient learning from neuromorphic sensors. We first introduce a spatial and temporal encoding scheme to map event-based neuromorphic data into high-dimensional space. Then, we leverage HDC mathematics to support learning and cognitive tasks over encoded data, such as information association and memorization. EventHD also provides a notion of confidence for each prediction, thus enabling self-learning from unlabeled data. We evaluate EventHD efficiency over data collected from Dynamic Vision Sensor (DVS) sensors. Our results indicate that EventHD can provide online learning and cognitive support while operating over raw DVS data without using the costly preprocessing step. In terms of efficiency, EventHD provides 14.2× faster and 19.8× higher energy efficiency than state-of-the-art learning algorithms while improving the computational robustness by 5.9×.

What Country, University, or Research Institute, Performed the Best on Covid-19 During the First Wave of the Pandemic?: Bibliometric analysis of scientific literature - analysing a 'snapshot in time' of the first wave of COVID-19

In this article, we conduct data mining and statistical analysis on the most effective countries, universities, and companies, based on their output (e.g., produced or collaborated) on COVID-19 during the first wave of the pandemic. Hence, the focus of this article is on the first wave of the pandemic. While in later stages of the pandemic, US and UK performed best in terms of vaccine production, the focus in this article is on the initial few months of the pandemic. The article presents findings from our analysing of all available records on COVID-19 from the Web of Science Core Collection. The results are compared with all available data records on pandemics and epidemics from 1900 to 2020. This has created interesting findings that are presented in the article with visualisation tools.

Supplementary information

The online version contains supplementary material available at 10.1007/s40745-022-00406-8.

Super-forecasting the 'technological singularity' risks from artificial intelligence

This article investigates cybersecurity (and risk) in the context of 'technological singularity' from artificial intelligence. The investigation constructs multiple risk forecasts that are synthesised in a new framework for counteracting risks from artificial intelligence (AI) itself. In other words, the research in this article is not just concerned with securing a system, but also analysing how the system responds when (internal and external) failure(s) and compromise(s) occur. This is an important methodological principle because not all systems can be secured, and totally securing a system is not feasible. Thus, we need to construct algorithms that will enable systems to continue operating even when parts of the system have been compromised. Furthermore, the article forecasts emerging cyber-risks from the integration of AI in cybersecurity. Based on the forecasts, the article is concentrated on creating synergies between the existing literature, the data sources identified in the survey, and forecasts. The forecasts are used to increase the feasibility of the overall research and enable the development of novel methodologies that uses AI to defend from cyber risks. The methodology is focused on addressing the risk of AI attacks, as well as to forecast the value of AI in defence and in the prevention of AI rogue devices acting independently.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12530-022-09431-7.

Memory-inspired spiking hyperdimensional network for robust online learning

Recently, brain-inspired computing models have shown great potential to outperform today's deep learning solutions in terms of robustness and energy efficiency. Particularly, Spiking Neural Networks (SNNs) and HyperDimensional Computing (HDC) have shown promising results in enabling efficient and robust cognitive learning. Despite the success, these two brain-inspired models have different strengths. While SNN mimics the physical properties of the human brain, HDC models the brain on a more abstract and functional level. Their design philosophies demonstrate complementary patterns that motivate their combination. With the help of the classical psychological model on memory, we propose SpikeHD, the first framework that fundamentally combines Spiking neural network and hyperdimensional computing. SpikeHD generates a scalable and strong cognitive learning system that better mimics brain functionality. SpikeHD exploits spiking neural networks to extract low-level features by preserving the spatial and temporal correlation of raw event-based spike data. Then, it utilizes HDC to operate over SNN output by mapping the signal into high-dimensional space, learning the abstract information, and classifying the data. Our extensive evaluation on a set of benchmark classification problems shows that SpikeHD provides the following benefit compared to SNN architecture: (1) significantly enhance learning capability by exploiting two-stage information processing, (2) enables substantial robustness to noise and failure, and (3) reduces the network size and required parameters to learn complex information.

Supervised learning model predicts protein adsorption to carbon nanotubes

Engineered nanoparticles are advantageous for biotechnology applications including biomolecular sensing and delivery. However, testing compatibility and function of nanotechnologies in biological systems requires a heuristic approach, where unpredictable protein corona formation prevents their effective implementation. We develop a random forest classifier trained with mass spectrometry data to identify proteins that adsorb to nanoparticles based solely on the protein sequence (78% accuracy, 70% precision). We model proteins that populate the corona of a single-walled carbon nanotube (SWCNT)–based nanosensor and study the relationship between the protein’s amino acid–based properties and binding capacity. Protein features associated with increased likelihood of SWCNT binding include high content of solvent-exposed glycines and nonsecondary structure–associated amino acids. To evaluate its predictive power, we apply the classifier to identify proteins with high binding affinity to SWCNTs, with experimental validation. The developed classifier provides a step toward undertaking the otherwise intractable problem of predicting protein-nanoparticle interactions.

Effects of lifelong testosterone exposure on health and disease using Mendelian randomization

Testosterone products are prescribed to males for a variety of possible health benefits, but causal effects are unclear. Evidence from randomized trials are difficult to obtain, particularly regarding effects on long-term or rare outcomes. Mendelian randomization analyses were performed to infer phenome-wide effects of free testosterone on 461 outcomes in 161,268 males from the UK Biobank study. Lifelong increased free testosterone had beneficial effects on increased bone mineral density, and decreased body fat; adverse effects on decreased HDL, and increased risks of prostate cancer, androgenic alopecia, spinal stenosis, and hypertension; and context-dependent effects on increased hematocrit and decreased C-reactive protein. No benefit was observed for type 2 diabetes, cardiovascular or cognitive outcomes. Mendelian randomization suggests benefits of long-term increased testosterone should be considered against adverse effects, notably increased prostate cancer and hypertension. Well-powered randomized trials are needed to conclusively address risks and benefits of testosterone treatment on these outcomes.

All-dielectric concentration of electromagnetic fields at the nanoscale: the role of photonic nanojets

The photonic nanojet (PNJ) is a narrow high-energy beam that was originally found on the back side of all-dielectric spherical structures. It is a unique type of energy concentration mode. The field of PNJs has experienced rapid growth in the past decade: nonspherical and even pixelized PNJ generators based on new physics and principles along with extended photonic applications from linear optics to nonlinear optics have driven the re-evaluation of the role of PNJs in optics and photonics. In this article, we give a comprehensive review for the emerging sub-topics in the past decade with a focus on two specific areas: (1) PNJ generators based on natural materials, artificial materials and nanostructures, and even programmable systems instead of conventional dielectric geometries such as microspheres, cubes, and trihedral prisms, and (2) the emerging novel applications in both linear and nonlinear optics that are built upon the specific features of PNJs. The extraordinary features of PNJs including subwavelength concentration of electromagnetic energy, high intensity focusing spot, and lower Joule heating as compared to plasmonic resonance systems, have made PNJs attractive to diverse fields spanning from optical imaging, nanofabrication, and integrated photonics to biosensing, optical tweezers, and disease diagnosis.