The energy demand of modern artificial intelligence (AI) has grown faster than the efficiency of the von Neumann hardware that runs it, renewing interest in neuromorphic computing as a brain-inspired alternative. This review surveys more than three decades of neuromorphic hardware, from Mead's foundational analog very-large-scale-integration (VLSI) work in 1990 to large-scale digital and mixed-signal platforms and emerging in-memory crossbar devices reported through 2026. We organize the field along four axes: the fundamentals of spike-based, event-driven computation; landmark hardware platforms including IBM TrueNorth, Intel Loihi and Loihi 2, SpiNNaker, BrainScaleS, Tianjic, and Darwin3; the compilation problem of mapping neural networks onto resource-constrained crossbar hardware; and the device technologies—memristors, resistive RAM (ReRAM), and related non-volatile memory—that promise sub-picojoule synaptic operations. Wherever quantitative comparisons are drawn, the figures are taken from the primary published literature and cited to their source: measured energies per synaptic operation ranging from 26 pJ for TrueNorth and 23.6 pJ for Loihi down to 5.47 pJ for Darwin3 and into the femtojoule regime for memristive devices, and quantization studies in which spiking networks retain accuracy within roughly one percentage point of full precision down to 2-bit weights. We synthesize these results, identify the recurring bottlenecks of device variability, communication overhead, and toolchain immaturity, and outline the open problems that will shape the next decade of the field.
Key Words: Artificial Intelligence (AI), Crossbar, In-memory Computing, Memristors, Neuromorphic Computing, ReRAM, Spiking Networks, VLSI.
Citation: R K Alajangi., “Neuromorphic Hardware for Energy-Efficient Artificial Intelligence: A Review of Architectures, Devices, and Compilation” Ci-STEM Journal of Digital Technologies and Expert Systems, Vol. 3(2), pp. 76-85, 2026, doi: 10.55306/CJDTES.2026.030201

 View Full Text (HTML)      Download Full Text

TColour filter array (CFA) demosaicking is the process by which a single-sensor digital camera reconstructs a full-resolution, three-channel colour image from a spatially sub-sampled mosaic of red, green and blue measurements. Because nearly every consumer camera, smartphone and machine-vision sensor in use today relies on some variant of a Bayer, Quad-Bayer or Nona-Bayer colour filter array, the quality of the demosaicking step places a hard ceiling on the achievable image quality of the entire imaging pipeline. Early work on this problem, spanning roughly 1987 to the mid-2000s, produced a rich family of model-based interpolation schemes: nearest-neighbour replication, bilinear and smooth-hue-transition interpolation, median-based and edge-sensing algorithms, block-matching, edge-directed interpolation, and directional-filtering-with-a-posteriori-decision schemes. These methods remain instructive because they encode, in closed analytic form, the core priors that all later data-driven methods must eventually learn: spectral correlation between colour channels, local edge orientation, and piecewise smoothness. Since approximately 2016, the field has been reshaped by deep learning. Convolutional networks, residual and iterative denoising cascades, attention and Transformer-based architectures, implicit neural representations, and, most recently, state-space (Mamba) and binarized hybrid networks have driven reported PSNR figures on standard benchmarks up by eight to fourteen decibels relative to bilinear interpolation, while simultaneously extending the task to jointly address denoising, non-Bayer sensor layouts (Quad-Bayer, Nona-Bayer, HybridEVS), and real-time mobile deployment constraints. This paper presents a comprehensive survey that (i) revisits the classical algorithmic foundations of demosaicking, (ii) systematically reviews the deep-learning literature from 2016 through 2026, organised by architectural paradigm, (iii) consolidates benchmark datasets and evaluation protocols used across this literature, (iv) compiles and compares quantitative results reported for representative algorithms on the Kodak and McMaster datasets, and (v) discusses open challenges — including generalisation across sensor patterns, joint demosaicking-denoising under real noise, and efficient on-device inference — together with promising future research directions such as unified multi-CFA models, diffusion-based priors, and self-supervised training regimes.
Key Words: Colour Filter Array, Demosaicking, Bayer Pattern, Quad-Bayer, Nona-Bayer, Convolutional Neural Networks, Vision Transformers, State-Space Models, Image Signal Processing, Joint Denoising-Demosaicking.
Citation: C. R. Rao et al., “From Classical Interpolation to Deep Neural Reconstruction: A Comprehensive Survey of Colour Filter Array Demosaicking Algorithms” Ci-STEM Journal of Digital Technologies and Expert Systems, Vol. 3(2), pp. 86-100, 2026, doi: 10.55306/CJDTES.2026.030202

 View Full Text (HTML)      Download Full Text

The introduction of ultra-dense 6G networks poses new and unprecedented routing challenges as it has dynamic topology, extensive device connectivity, and high latency demands. Conventional routing schemes and the classical metaheuristic methods are unable to adjust effectively to extremely dynamic environments and cannot satisfy ultra-low latency requirements. The gap in research is that no adaptive, latency-aware routing systems exist that can effectively search large spaces of solutions and be stable in ultra-dense conditions. To handle this, a Quantum-Inspired Metaheuristic Routing (QIMR) algorithm is suggested that combines quantum annealing algorithms with hybrid metaheuristic optimization, to help find intelligent routes. The model involves quantum state representation, probabilistic transition modeling, and latency-sensitive fitness evaluation. Simulated 6G ultra-dense network environments with realistic traffic conditions and benchmark datasets based on network simulators are used to conduct experiments. The proposed model demonstrates an average routing efficiency of 97.62, a ratio of 96.85 packet delivery and a 31 reduction in latency than baseline approaches. These findings indicate the effectiveness of the suggested framework in addressing complicated routing optimization issues in 6G settings of the future.
Key Words: 6G Networks, Latency Optimization, Metaheuristic Routing, Network Slicing, Quantum-Inspired Optimization, Ultra-Dense Networks.
Citation: C. R. Rao et al., “Quantum-Inspired Metaheuristic Routing for Ultra-Dense 6G Networks”, Ci-STEM Journal of Digital Technologies and Expert Systems, Vol. 3(2), pp. 101-108, 2026, doi: 10.55306/CJDTES.2026.030203

 View Full Text (HTML)      Download Full Text

Cyber-attacks on computer networks are evolving, and signature-based defence systems are struggling to keep pace. One of the more pragmatic answers to this problem has been machine learning, not because it replaces a security analyst, but because it can sift through traffic much faster than a human team while catching patterns that static rules miss. This paper investigates the performance of four supervised classifiers, namely Support Vector Machine (SVM), Decision Tree, Random Forest and an Artificial Neural Network (ANN) on the NSL-KDD dataset, which is an improved, de-duplicated version of the older KDD Cup 1999 data. All four models were run through the same pre-processing, feature engineering and validation steps so there is no skew in the comparison due to inconsistent handling of the data. It is important to note upfront how much reported IDS accuracy depends on the dataset and protocol used. For example, work on evaluating SVM on CICIDS2017 has reported precision as high as 97.80% on one attack category and as low as 69.79% on another, which is a widespread for the same algorithm. In our own runs on NSL-KDD Random Forest scored the best with 99.6% accuracy while ANN came close at 99.11% and SVM and Decision Tree stood behind at 93.29% and 83.4% respectively. The pattern suggests that ensemble tree methods are well suited for the discrete, high-dimensional structure of network connection records. The paper also puts these numbers alongside recent feature-selection- and deep-learning-based intrusion detection works to see where the classical approach still holds its ground and where it doesn’t.
Key Words: Artificial neural network, Cybercrime, Decision tree, Intrusion detection system (IDS), Machine learning, Network security, NSL-KDD, Random forest, Support vector machine.
Citation: Sri Nagesh A et al., “Quantum Inspired Metaheuristic Routing for Ultra-Dense 6G Networks”, Ci-STEM Journal of Digital Technologies and Expert Systems, Vol. 3(2), pp. 109-124, 2026, doi: 10.55306/CJDTES.2026.030204

 View Full Text (HTML)      Download Full Text