CyberSec Research

In this paper we argue that ReLU networks learn an implicit linear model we can actually tap into. We describe that alleged model formally and show that we can approximately pull its decision boundary back to the input space with certain simple modification to the backward pass. The resulting gradients (called excitation pullbacks) reveal high-resolution input- and target-specific features of remarkable perceptual alignment on a number of popular ImageNet-pretrained deep architectures. This strongly suggests that neural networks do, in fact, rely on learned interpretable patterns that can be recovered after training. Thus, our findings may have profound implications for knowledge discovery and the development of dependable artificial systems.

We present a vectorial extension of the Hopfield associative memory model inspired by the theory of amorphous solids, where binary neural states are replaced by unit vectors $\mathbf{s}_i \in \mathbb{R}^3$ on the sphere $S^2$. The generalized Hebbian learning rule creates a block-structured weight matrix through outer products of stored pattern vectors, analogous to the Hessian matrix structure in amorphous solids. We demonstrate that this model exhibits quantifiable structural properties characteristic of disordered materials: energy landscapes with deep minima for stored patterns versus random configurations (energy gaps $\sim 7$ units), strongly anisotropic correlations encoded in the weight matrix (anisotropy ratios $\sim 10^2$), and order-disorder transitions controlled by the pattern density $\gamma = P/(N \cdot d)$. The enhanced memory capacity ($\gamma_c \approx 0.55$ for a fully-connected network) compared to binary networks ($\gamma_c \approx 0.138$) and the emergence of orientational correlations establish connections between associative memory mechanisms and amorphous solid physics, particularly in systems with continuous orientational degrees of freedom. We also unveil the scaling with the coordination number $Z$ of the memory capacity: $\gamma_c \sim (Z-6)$ from the isostatic point $Z_c =6$ of the 3D elastic network, which closely mirrors the scaling of the shear modulus $G \sim (Z-6)$ in 3D central-force spring networks.

The Nearest-Better Network (NBN) is a powerful method to visualize sampled data for continuous optimization problems while preserving multiple landscape features. However, the calculation of NBN is very time-consuming, and the extension of the method to combinatorial optimization problems is challenging but very important for analyzing the algorithm's behavior. This paper provides a straightforward theoretical derivation showing that the NBN network essentially functions as the maximum probability transition network for algorithms. This paper also presents an efficient NBN computation method with logarithmic linear time complexity to address the time-consuming issue. By applying this efficient NBN algorithm to the OneMax problem and the Traveling Salesman Problem (TSP), we have made several remarkable discoveries for the first time: The fitness landscape of OneMax exhibits neutrality, ruggedness, and modality features. The primary challenges of TSP problems are ruggedness, modality, and deception. Two state-of-the-art TSP algorithms (i.e., EAX and LKH) have limitations when addressing challenges related to modality and deception, respectively. LKH, based on local search operators, fails when there are deceptive solutions near global optima. EAX, which is based on a single population, can efficiently maintain diversity. However, when multiple attraction basins exist, EAX retains individuals within multiple basins simultaneously, reducing inter-basin interaction efficiency and leading to algorithm's stagnation.

We propose a biologically inspired model of spiking neurons based on the dynamics of a damped, driven pendulum. Unlike traditional models such as the Leaky Integrate-and-Fire (LIF) neurons, the pendulum neuron incorporates second-order, nonlinear dynamics that naturally give rise to oscillatory behavior and phase-based spike encoding. This model captures richer temporal features and supports timing-sensitive computations critical for sequence processing and symbolic learning. We present an analysis of single-neuron dynamics and extend the model to multi-neuron layers governed by Spike-Timing Dependent Plasticity (STDP) learning rules. We demonstrate practical implementation with python code and with the Brian2 spiking neural simulator, and outline a methodology for deploying the model on neuromorphic hardware platforms, using an approximation of the second-order equations. This framework offers a foundation for developing energy-efficient neural systems for neuromorphic computing and sequential cognition tasks.

Service Function Chains (SFCs) are one of the key enablers in providing programmable computer networks, paving the way for network autonomy. However, this also introduces new challenges, such as resource allocation and optimisation related to their operation, requiring new algorithms to address these challenges. Various tools have been used in the literature to evaluate these algorithms. However, these tools suffer from inaccuracy, low fidelity, unscalability, inflexibility, or additional code requirements. This paper introduces an emulator based on Mininet and Docker for SFCs called OpenRASE. The goal of OpenRASE is to enable the exploration of resource allocation algorithms for SFCs in a dynamic setting, allowing real CPU usage and latency to be measured. We describe the design and implementation of OpenRASE and discuss its characteristics. We also experimentally evaluate two different algorithms to address the SFC resource allocation challenge, including an online Genetic Algorithm, using OpenRASE to show its effectiveness and practicality for dynamic network conditions.

The capacitated arc routing problem with time-dependent service costs (CARPTDSC) is a challenging combinatorial optimization problem that arises from winter gritting applications. CARPTDSC has two main challenges about time consumption. First, it is an NP-hard problem. Second, the time-dependent service costs of tasks require frequent evaluations during the search process, significantly increasing computational effort. These challenges make it difficult for existing algorithms to perform efficient searches, often resulting in limited efficiency. To address these issues, this paper proposes a knowledge-guided memetic algorithm with golden section search and negatively correlated search (KGMA-GN), where two knowledge-guided strategies are introduced to improve search efficiency. First, a knowledge-guided initialization strategy (KGIS) is proposed to generate high-quality initial solutions to speed up convergence. Second, a knowledge-guided small-step-size local search strategy (KGSLSS) is proposed to filter out invalid moves, thereby reducing unnecessary evaluations and saving the computation time. Experimental results on five benchmark test sets, including both small- and larger-scale instances, demonstrate that KGMA-GN achieves higher search efficiency than the state-of-the-art methods. Moreover, the ablation study further confirms that the knowledge-guided local search operators in KGSLSS can significantly reduce runtime compared to traditional operators, especially for the knowledge-guided swap operator, which achieves more than a tenfold improvement in speed.

Activation functions are critical components in deep neural networks, directly influencing gradient flow, training stability, and model performance. Traditional functions like ReLU suffer from dead neuron problems, while sigmoid and tanh exhibit vanishing gradient issues. We introduce two novel hybrid activation functions: S3 (Sigmoid-Softsign) and its improved version S4 (smoothed S3). S3 combines sigmoid for negative inputs with softsign for positive inputs, while S4 employs a smooth transition mechanism controlled by a steepness parameter k. We conducted comprehensive experiments across binary classification, multi-class classification, and regression tasks using three different neural network architectures. S4 demonstrated superior performance compared to nine baseline activation functions, achieving 97.4% accuracy on MNIST, 96.0% on Iris classification, and 18.7 MSE on Boston Housing regression. The function exhibited faster convergence (-19 for ReLU) and maintained stable gradient flow across network depths. Comparative analysis revealed S4's gradient range of [0.24, 0.59] compared to ReLU's 18% dead neurons in deep networks. The S4 activation function addresses key limitations of existing functions through its hybrid design and smooth transition mechanism. The tunable parameter k allows adaptation to different tasks and network depths, making S4 a versatile choice for deep learning applications. These findings suggest that hybrid activation functions represent a promising direction for improving neural network training dynamics.

Despite success across diverse tasks, current artificial recurrent network architectures rely primarily on implicit hidden-state memories, limiting their interpretability and ability to model long-range dependencies. In contrast, biological neural systems employ explicit, associative memory traces (i.e., engrams) strengthened through Hebbian synaptic plasticity and activated sparsely during recall. Motivated by these neurobiological insights, we introduce the Engram Neural Network (ENN), a novel recurrent architecture incorporating an explicit, differentiable memory matrix with Hebbian plasticity and sparse, attention-driven retrieval mechanisms. The ENN explicitly models memory formation and recall through dynamic Hebbian traces, improving transparency and interpretability compared to conventional RNN variants. We evaluate the ENN architecture on three canonical benchmarks: MNIST digit classification, CIFAR-10 image sequence modeling, and WikiText-103 language modeling. Our empirical results demonstrate that the ENN achieves accuracy and generalization performance broadly comparable to classical RNN, GRU, and LSTM architectures, with all models converging to similar accuracy and perplexity on the large-scale WikiText-103 task. At the same time, the ENN offers significant enhancements in interpretability through observable memory dynamics. Hebbian trace visualizations further reveal biologically plausible, structured memory formation processes, validating the potential of neuroscience-inspired mechanisms to inform the development of more interpretable and robust deep learning models.

Reservoir Computing is a machine learning approach that uses the rich repertoire of complex system dynamics for function approximation. Current approaches to reservoir computing use a network of coupled integrating neurons that require a steady current to maintain activity. Here, we introduce a small world graph of differentiating neurons that are active only when there are changes in input as an alternative to integrating neurons as a reservoir computing substrate. We find the coupling strength and network topology that enable these small world networks to function as an effective reservoir. We demonstrate the efficacy of these networks in the MNIST digit recognition task, achieving comparable performance of 90.65% to existing reservoir computing approaches. The findings suggest that differentiating neurons can be a potential alternative to integrating neurons and can provide a sustainable future alternative for power-hungry AI applications.

Memristor-based Spiking Neural Networks (SNNs) with temporal spike encoding enable ultra-low-energy computation, making them ideal for battery-powered intelligent devices. This paper presents a circuit-level memristive spiking neural network (SNN) architecture trained using a proposed novel supervised in-situ learning algorithm inspired by spike-timing-dependent plasticity (STDP). The proposed architecture efficiently implements lateral inhibition and the refractory period, eliminating the need for external microcontrollers or ancillary control hardware. All synapses of the winning neurons are updated in parallel, enhancing training efficiency. The modular design ensures scalability with respect to input data dimensions and output class count. The SNN is evaluated in LTspice for pattern recognition (using 5x3 binary images) and classification tasks using the Iris and Breast Cancer Wisconsin (BCW) datasets. During testing, the system achieved perfect pattern recognition and high classification accuracies of 99.11\% (Iris) and 97.9\% (BCW). Additionally, it has demonstrated robustness, maintaining an average recognition rate of 93.4\% under 20\% input noise. The impact of stuck-at-conductance faults and memristor device variations was also analyzed.

Multi-objective combinatorial optimization problems (MOCOP) frequently arise in practical applications that require the simultaneous optimization of conflicting objectives. Although traditional evolutionary algorithms can be effective, they typically depend on domain knowledge and repeated parameter tuning, limiting flexibility when applied to unseen MOCOP instances. Recently, integration of Large Language Models (LLMs) into evolutionary computation has opened new avenues for automatic heuristic generation, using their advanced language understanding and code synthesis capabilities. Nevertheless, most existing approaches predominantly focus on single-objective tasks, often neglecting key considerations such as runtime efficiency and heuristic diversity in multi-objective settings. To bridge this gap, we introduce Multi-heuristics for MOCOP via Pareto-Grid-guided Evolution of LLMs (MPaGE), a novel enhancement of the Simple Evolutionary Multiobjective Optimization (SEMO) framework that leverages LLMs and Pareto Front Grid (PFG) technique. By partitioning the objective space into grids and retaining top-performing candidates to guide heuristic generation, MPaGE utilizes LLMs to prioritize heuristics with semantically distinct logical structures during variation, thus promoting diversity and mitigating redundancy within the population. Through extensive evaluations, MPaGE demonstrates superior performance over existing LLM-based frameworks, and achieves competitive results to traditional Multi-objective evolutionary algorithms (MOEAs), with significantly faster runtime. Our code is available at: https://github.com/langkhachhoha/MPaGE.

This paper preliminarily investigates the duality between flow matching in generative models and particle swarm optimization (PSO) in evolutionary computation. Through theoretical analysis, we reveal the intrinsic connections between these two approaches in terms of their mathematical formulations and optimization mechanisms: the vector field learning in flow matching shares similar mathematical expressions with the velocity update rules in PSO; both methods follow the fundamental framework of progressive evolution from initial to target distributions; and both can be formulated as dynamical systems governed by ordinary differential equations. Our study demonstrates that flow matching can be viewed as a continuous generalization of PSO, while PSO provides a discrete implementation of swarm intelligence principles. This duality understanding establishes a theoretical foundation for developing novel hybrid algorithms and creates a unified framework for analyzing both methods. Although this paper only presents preliminary discussions, the revealed correspondences suggest several promising research directions, including improving swarm intelligence algorithms based on flow matching principles and enhancing generative models using swarm intelligence concepts.

Spiking neural networks possess the advantage of low energy consumption due to their event-driven nature. Compared with binary spike outputs, their inherent floating-point dynamics are more worthy of attention. The threshold level and reset mode of neurons play a crucial role in determining the number and timing of spikes. The existing hard reset method causes information loss, while the improved soft reset method adopts a uniform treatment for neurons. In response to this, this paper designs an adaptive reset neuron, establishing the correlation between input, output and reset, and integrating a simple yet effective threshold adjustment strategy. It achieves excellent performance on various datasets while maintaining the advantage of low energy consumption.

Stock trading has always been a challenging task due to the highly volatile nature of the stock market. Making sound trading decisions to generate profit is particularly difficult under such conditions. To address this, we propose four novel loss functions to drive decision-making for a portfolio of stocks. These functions account for the potential profits or losses based with respect to buying or shorting respective stocks, enabling potentially any artificial neural network to directly learn an effective trading strategy. Despite the high volatility in stock market fluctuations over time, training time-series models such as transformers on these loss functions resulted in trading strategies that generated significant profits on a portfolio of 50 different S&P 500 company stocks as compared to a benchmark reinforcment learning techniques and a baseline buy and hold method. As an example, using 2021, 2022 and 2023 as three test periods, the Crossformer model adapted with our best loss function was most consistent, resulting in returns of 51.42%, 51.04% and 48.62% respectively. In comparison, the best performing state-of-the-art reinforcement learning methods, PPO and DDPG, only delivered maximum profits of around 41%, 2.81% and 41.58% for the same periods. The code is available at https://anonymous.4open.science/r/bandit-stock-trading-58C8/README.md.

Feed-forward neural networks (FFNNs) are vulnerable to input noise, reducing prediction performance. Existing regularization methods like dropout often alter network architecture or overlook neuron interactions. This study aims to enhance FFNN noise robustness by modifying backpropagation, interpreted as a multi-agent game, and exploring controlled target variable noising. Our "gradient dropout" selectively nullifies hidden layer neuron gradients with probability 1 - p during backpropagation, while keeping forward passes active. This is framed within compositional game theory. Additionally, target variables were perturbed with white noise or stable distributions. Experiments on ten diverse tabular datasets show varying impacts: improvement or diminishing of robustness and accuracy, depending on dataset and hyperparameters. Notably, on regression tasks, gradient dropout (p = 0.9) combined with stable distribution target noising significantly increased input noise robustness, evidenced by flatter MSE curves and more stable SMAPE values. These results highlight the method's potential, underscore the critical role of adaptive parameter tuning, and open new avenues for analyzing neural networks as complex adaptive systems exhibiting emergent behavior within a game-theoretic framework.

We introduce Pareto-NRPA, a new Monte-Carlo algorithm designed for multi-objective optimization problems over discrete search spaces. Extending the Nested Rollout Policy Adaptation (NRPA) algorithm originally formulated for single-objective problems, Pareto-NRPA generalizes the nested search and policy update mechanism to multi-objective optimization. The algorithm uses a set of policies to concurrently explore different regions of the solution space and maintains non-dominated fronts at each level of search. Policy adaptation is performed with respect to the diversity and isolation of sequences within the Pareto front. We benchmark Pareto-NRPA on two classes of problems: a novel bi-objective variant of the Traveling Salesman Problem with Time Windows problem (MO-TSPTW), and a neural architecture search task on well-known benchmarks. Results demonstrate that Pareto-NRPA achieves competitive performance against state-of-the-art multi-objective algorithms, both in terms of convergence and diversity of solutions. Particularly, Pareto-NRPA strongly outperforms state-of-the-art evolutionary multi-objective algorithms on constrained search spaces. To our knowledge, this work constitutes the first adaptation of NRPA to the multi-objective setting.

Recommender systems, especially those based on graph neural networks (GNNs), have achieved remarkable success in capturing user-item interaction patterns. However, they remain susceptible to popularity bias--the tendency to over-recommend popular items--resulting in reduced content diversity and compromised fairness. In this paper, we propose PBiLoss, a novel regularization-based loss function designed to counteract popularity bias in graph-based recommender models explicitly. PBiLoss augments traditional training objectives by penalizing the model's inclination toward popular items, thereby encouraging the recommendation of less popular but potentially more personalized content. We introduce two sampling strategies: Popular Positive (PopPos) and Popular Negative (PopNeg), which respectively modulate the contribution of the positive and negative popular items during training. We further explore two methods to distinguish popular items: one based on a fixed popularity threshold and another without any threshold, making the approach flexible and adaptive. Our proposed method is model-agnostic and can be seamlessly integrated into state-of-the-art graph-based frameworks such as LightGCN and its variants. Comprehensive experiments across multiple real-world datasets demonstrate that PBiLoss significantly improves fairness, as demonstrated by reductions in the Popularity-Rank Correlation for Users (PRU) and Popularity-Rank Correlation for Items (PRI), while maintaining or even enhancing standard recommendation accuracy and ranking metrics. These results highlight the effectiveness of directly embedding fairness objectives into the optimization process, providing a practical and scalable solution for balancing accuracy and equitable content exposure in modern recommender systems.

Systematic compositional generalization - constructing and understanding novel combinations of known building blocks - remains a core challenge for AI systems. Human cognition achieves this flexibility via the interplay of the hippocampus (HPC) and prefrontal cortex (PFC): the hippocampus rapidly encodes episodes, and the prefrontal cortex consolidates them into reusable schemas for reasoning. Drawing on these insights, we present MIRAGE (Meta-Inference with Rules and Abstractions from Generalized Experience), a framework that achieves systematic generalization on compositional tasks. MIRAGE has two interacting modules mirroring the brain's deliberative HPC-PFC loop and intuitive neocortical pattern recognition. (1) The meta-trained Transformer Neural Decomposer, paralleling neocortical "System 1" computation, is trained on a task-agnostic stream of randomly sampled compositional grammars and applies one decomposition step per pass, with successive passes iteratively refining the sequence representation. (2) The Schema Engine, analogous to the HPC-PFC "System 2" loop, dynamically extracts, ranks, and applies reusable schemas, storing variable bindings in episodic memory and expanding them when needed. By explicitly equipping the Transformer component of MIRAGE with actively managed schematic structures, our model performs systematic compositional operations through explicit schema application and transformation, relying solely on frozen weights when solving entirely novel tasks. This approach demonstrates systematic compositional generalization on the SCAN benchmark, achieving > 99% accuracy on all task splits with only 1.19M parameters in the transformer module. Ablation studies confirm that MIRAGE's systematicity critically depends on the quality of extracted schemas and the model's iterative refinement process.