arXiv subjects

Van-Tam Nguyen

Publications and source records attributed to Van-Tam Nguyen.

At least 19 recordsLinked to original sources

arXiv2026

Predicting Only from Selected Evidence: A Tempered Product-of-Experts Bottleneck for Auditable EEG Diagnosis

Pretrained EEG backbones improve transfer performance, but downstream diagnosis heads remain hard to audit: predictions are made from unrestricted hidden states, whereas explanations are usually produced only after the decision. We introduce tPoE-EIB, an evidence-information bottleneck head for adapting EEG backbones under an evidence-only prediction constraint. tPoE-EIB selects temporal and channel evidence, maps the selected summaries to Gaussian experts over a shared latent variable, and fuses them with a tempered product-of-experts posterior. The classifier observes only this latent, so the decision path is explicit and rate-limited by the expected posterior KL. This gives a tractable supervised objective with an information-rate penalty, while the closed-form tempered posterior mitigates overconfident fusion from correlated evidence axes. We evaluate tPoE-EIB on pretrained EEG foundation-model backbones across six diagnosis settings: event-type classification, abnormality detection, seizure detection, cognitive-decline staging, depression screening, and cerebrovascular-disease classification. The evaluation spans public benchmarks and in-house clinical cohorts, binary screening and fine-grained staging, and sparse and dense montages. tPoE-EIB preserves competitive balanced accuracy and improves over representative post-hoc explanations on selection-faithfulness audits, including insertion-deletion and gate-causality tests. Its structured posterior further enables integration-faithfulness audits, including expert-drop, posterior-reliance, and expert-disagreement tests. Overall, these results suggest that evidence-only, rate-limited fusion is a practical route to auditable diagnosis on top of frozen EEG foundation backbones.

eess.SP↗

arXiv2026

How Benchmarks and Evaluation Protocols Shape Conclusions in Provenance-Based Intrusion Detection

Provenance-based intrusion detection systems (PIDS) frequently report strong performance, but the conclusions drawn from these results can be highly sensitive to benchmarking choices and evaluation protocols. We investigate this dependency by re-evaluating representative PIDS on public datasets that meet our audit, labeling, and calibration requirements. Focusing primarily on the audited DARPA TC E3 datasets, we apply a unified protocol with temporally separated test periods and validation-only checkpoint selection and threshold calibration, and ask which architectural claims are empirically supported. We find that alerting success and investigation utility can diverge sharply, as several systems surface attacks without providing enough process-level context to support forensic investigation. Across the four primary datasets, a simple allowlist built from executable names and paths observed during training matches or exceeds the selected learned baselines on key operating-point metrics, showing that comparable performance on these metrics is achievable using lexical novelty alone. Quantifying semantic signal quality through feature completeness and field entropy helps explain why several audited E3 datasets support alerting performance without reliably separating model architectures. In contrast, Theia provides the richest semantic signal and shows the clearest improvements in ranking and node-level recovery for our reference model. Overall, these findings reinforce the importance of interpreting architectural claims in PIDS together with the benchmark properties and evaluation protocol that produced them.

cs.CR↗

arXiv2026

TORINO: Token Reduction via Interpretable Concept Overlap in Vision-Language Models

Vision-Language Models (VLMs) have demonstrated impressive capabilities across different tasks, but their computational cost is dominated by the large number of visual tokens fed to the language model. Existing token reduction methods rely on attention-based scores or pairwise similarity, without an explicit semantic representation of each token. We introduce TORINO (TOken Reduction via Interpretable coNcept Overlap), a plug-and-play framework for adaptive visual token reduction in VLMs that requires no fine-tuning of the underlying model. TORINO leverages Sparse Autoencoders (SAEs) to project visual tokens into an interpretable latent space where token relationships can be analyzed through shared concept activations. Specifically, we define concept overlap as the degree of agreement between active SAE latents and use it to group tokens that share semantic content. Reduction within each group is then performed by either pruning or merging, providing a unified framework that preserves semantically important visual information while removing redundancy. Unlike fixed-budget approaches, TORINO dynamically adapts the reduction rate to input complexity, allowing different images to retain different numbers of tokens. Experiments across multiple vision-language benchmarks show that TORINO achieves favorable efficiency-accuracy trade-offs, reducing the number of visual tokens with minimal performance loss.

cs.CV↗

arXiv2026

Efficient Resource-Constrained Training of Transformers via Subspace Optimization

As AI increasingly shapes daily life, energy consumption and data privacy have become pressing concerns. On-device learning trains models directly on edge devices, cutting energy consumption and safeguarding data privacy. However, the expanding scale of modern neural networks creates a major obstacle for on-device training. Although prior work has concentrated on compact convolutional architectures, we instead apply subspace-based training to transformer models. Motivated by the idea that a model's essential information lies in a fixed subspace, we introduce Weight-Activation Subspace Iteration (WASI), a method that mitigates the memory bottleneck of backpropagation and boosts inference efficiency in transformer models by restricting training to this subspace. Our results demonstrate that WASI maintains accuracy comparable to vanilla training while reducing memory usage by up to $62\times$ and computational cost (FLOPs) by up to $2\times$. On a Raspberry Pi 5, WASI achieves roughly $1.4\times$ faster training and inference than vanilla training. The code is available at https://github.com/Le-TrungNguyen/ICLR2026-WASI.git.

cs.LG↗

arXiv2026

Layer Collapse Can be Induced by Unstructured Pruning

Unstructured pruning is a popular compression method for efficiently reducing model parameters. However, while it effectively decreases the number of parameters, it is commonly believed that unstructured pruning cannot shorten the computational critical path, i.e., the maximum number of layers traversed during forward propagation. In this paper, we study when and how unstructured pruning can yield structural effects. For rectifier-activated networks, we introduce the notion of neuron entropy, which quantifies the degree of nonlinearity utilization. We show that magnitude-based pruning naturally lowers this entropy, sometimes down to zero-entropy layers that become linearizable and can thus be removed. Building on this insight, we propose a method that leverages "unstructured" pruning to favor sparsity in low-entropy layers, enabling their complete removal. We validate the phenomenon across CNNs, Vision Transformers, and NLP models: unstructured pruning can induce effective layer removal with little or no performance degradation in over-parameterized networks.

cs.LG↗

arXiv2026

Study of Training Dynamics for Memory-Constrained Fine-Tuning

Memory-efficient training of deep neural networks has become increasingly important as models grow larger while deployment environments impose strict resource constraints. We propose TraDy, a novel transfer learning scheme leveraging two key insights: layer importance for updates is architecture-dependent and determinable a priori, while dynamic stochastic channel selection provides superior gradient approximation compared to static approaches. We introduce a dynamic channel selection approach that stochastically resamples channels between epochs within preselected layers. Extensive experiments demonstrate TraDy achieves state-of-the-art performance across various downstream tasks and architectures while maintaining strict memory constraints, achieving up to 99% activation sparsity, 95% weight derivative sparsity, and 97% reduction in FLOPs for weight derivative computation.

cs.LG↗

arXiv2026

HFMCA: Orthonormal Feature Learning for EEG-based Brain Decoding

Electroencephalography (EEG) analysis is critical for brain-computer interfaces and neuroscience, but the intrinsic noise and high dimensionality of EEG signals hinder effective feature learning. We propose a self-supervised framework based on the Hierarchical Functional Maximal Correlation Algorithm (HFMCA), which learns orthonormal EEG representations by enforcing feature decorrelation and reducing redundancy. This design enables robust capture of essential brain dynamics for various EEG recognition tasks. We validate HFMCA on two benchmark datasets, SEED and BCIC-2A, where pretraining with HFMCA consistently outperforms competitive self-supervised baselines, achieving notable gains in classification accuracy. Across diverse EEG tasks, our method demonstrates superior cross-subject generalization under leave-one-subject-out validation, advancing state-of-the-art by 2.71\% on SEED emotion recognition and 2.57\% on BCIC-2A motor imagery classification.

eess.SP↗

arXiv2025

Memory Constrained Dynamic Subnetwork Update for Transfer Learning

On-device neural network training faces critical memory constraints that limit the adaptation of pre-trained models to downstream tasks. We present MeDyate, a theoretically-grounded framework for memory-constrained dynamic subnetwork adaptation. Our approach introduces two key innovations: LaRa (Layer Ranking), an improved layer importance metric that enables principled layer pre-selection, and a dynamic channel sampling strategy that exploits the temporal stability of channel importance distributions during fine-tuning. MeDyate dynamically resamples channels between epochs according to importance-weighted probabilities, ensuring comprehensive parameter space exploration while respecting strict memory budgets. Extensive evaluation across a large panel of tasks and architectures demonstrates that MeDyate achieves state-of-the-art performance under extreme memory constraints, consistently outperforming existing static and dynamic approaches while maintaining high computational efficiency. Our method represents a significant step towards enabling efficient on-device learning by demonstrating effective fine-tuning with memory budgets as low as a few hundred kB of RAM.

cs.LG↗

arXiv2025

Self-Supervised Learning of Graph Representations for Network Intrusion Detection

Detecting intrusions in network traffic is a challenging task, particularly under limited supervision and constantly evolving attack patterns. While recent works have leveraged graph neural networks for network intrusion detection, they often decouple representation learning from anomaly detection, limiting the utility of the embeddings for identifying attacks. We propose GraphIDS, a self-supervised intrusion detection model that unifies these two stages by learning local graph representations of normal communication patterns through a masked autoencoder. An inductive graph neural network embeds each flow with its local topological context to capture typical network behavior, while a Transformer-based encoder-decoder reconstructs these embeddings, implicitly learning global co-occurrence patterns via self-attention without requiring explicit positional information. During inference, flows with unusually high reconstruction errors are flagged as potential intrusions. This end-to-end framework ensures that embeddings are directly optimized for the downstream task, facilitating the recognition of malicious traffic. On diverse NetFlow benchmarks, GraphIDS achieves strong performance, reaching up to 99.98% PR-AUC and 99.61% macro F1-score.

cs.LG↗

arXiv2025

Beyond Low-rank Decomposition: A Shortcut Approach for Efficient On-Device Learning

On-device learning has emerged as a promising direction for AI development, particularly because of its potential to reduce latency issues and mitigate privacy risks associated with device-server communication, while improving energy efficiency. Despite these advantages, significant memory and computational constraints still represent major challenges for its deployment. Drawing on previous studies on low-rank decomposition methods that address activation memory bottlenecks in backpropagation, we propose a novel shortcut approach as an alternative. Our analysis and experiments demonstrate that our method can reduce activation memory usage, even up to $120.09\times$ compared to vanilla training, while also reducing overall training FLOPs up to $1.86\times$ when evaluated on traditional benchmarks.

cs.LG↗

arXiv2025

An FPGA Compiler for On-the-Fly Adaptive CNN Deployment and Reconfiguration

We introduce ForgeMorph, a full-stack compiler for adaptive CNN deployment on FPGAs, combining design-time optimization with runtime reconfigurability. At compile time, the NeuroForge engine performs constraint-driven design space exploration, generating RTL mappings that are Pareto-optimal with respect to user-defined latency and resource budgets. Unlike existing FPGA compilers, which rely on static scheduling and manual tuning, NeuroForge leverages analytical performance models and multi-objective genetic algorithms to efficiently search large configuration spaces and propose highly optimized hardware implementations. At runtime, the NeuroMorph module enables dynamic reconfiguration of network width and depth without requiring redeployment. This is made possible by a novel training strategy, DistillCycle, which jointly trains the full model and its subnetworks using hierarchical knowledge distillation. As a result, each execution path maintains accuracy even under aggressive resource and power constraints. We demonstrate Forge-Morph on the Zynq-7100 using custom and benchmark models including MobileNetV2, ResNet-50, SqueezeNet, and YOLOv5. The system achieves up to 50x latency reduction and 32% lower power consumption at runtime, while matching or exceeding the efficiency of state-of-the-art compilers. ForgeMorph offers a unified solution for deployment scenarios that demand flexibility, performance, and hardware efficiency

cs.AR↗

arXiv2024

Till the Layers Collapse: Compressing a Deep Neural Network through the Lenses of Batch Normalization Layers

Today, deep neural networks are widely used since they can handle a variety of complex tasks. Their generality makes them very powerful tools in modern technology. However, deep neural networks are often overparameterized. The usage of these large models consumes a lot of computation resources. In this paper, we introduce a method called \textbf{T}ill the \textbf{L}ayers \textbf{C}ollapse (TLC), which compresses deep neural networks through the lenses of batch normalization layers. By reducing the depth of these networks, our method decreases deep neural networks' computational requirements and overall latency. We validate our method on popular models such as Swin-T, MobileNet-V2, and RoBERTa, across both image classification and natural language processing (NLP) tasks.

cs.LG↗

arXiv2024

Activation Map Compression through Tensor Decomposition for Deep Learning

Internet of Things and Deep Learning are synergetically and exponentially growing industrial fields with a massive call for their unification into a common framework called Edge AI. While on-device inference is a well-explored topic in recent research, backpropagation remains an open challenge due to its prohibitive computational and memory costs compared to the extreme resource constraints of embedded devices. Drawing on tensor decomposition research, we tackle the main bottleneck of backpropagation, namely the memory footprint of activation map storage. We investigate and compare the effects of activation compression using Singular Value Decomposition and its tensor variant, High-Order Singular Value Decomposition. The application of low-order decomposition results in considerable memory savings while preserving the features essential for learning, and also offers theoretical guarantees to convergence. Experimental results obtained on main-stream architectures and tasks demonstrate Pareto-superiority over other state-of-the-art solutions, in terms of the trade-off between generalization and memory footprint.

cs.LG↗

arXiv2024

Memory-Optimized Once-For-All Network

Deploying Deep Neural Networks (DNNs) on different hardware platforms is challenging due to varying resource constraints. Besides handcrafted approaches aiming at making deep models hardware-friendly, Neural Architectures Search is rising as a toolbox to craft more efficient DNNs without sacrificing performance. Among these, the Once-For-All (OFA) approach offers a solution by allowing the sampling of well-performing sub-networks from a single supernet -- this leads to evident advantages in terms of computation. However, OFA does not fully utilize the potential memory capacity of the target device, focusing instead on limiting maximum memory usage per layer. This leaves room for an unexploited potential in terms of model generalizability. In this paper, we introduce a Memory-Optimized OFA (MOOFA) supernet, designed to enhance DNN deployment on resource-limited devices by maximizing memory usage (and for instance, features diversity) across different configurations. Tested on ImageNet, our MOOFA supernet demonstrates improvements in memory exploitation and model accuracy compared to the original OFA supernet. Our code is available at https://github.com/MaximeGirard/memory-optimized-once-for-all.

cs.LG↗

arXiv2024

AI-Driven Intrusion Detection Systems (IDS) on the ROAD Dataset: A Comparative Analysis for Automotive Controller Area Network (CAN)

The integration of digital devices in modern vehicles has revolutionized automotive technology, enhancing safety and the overall driving experience. The Controller Area Network (CAN) bus is a central system for managing in-vehicle communication between the electronic control units (ECUs). However, the CAN protocol poses security challenges due to inherent vulnerabilities, lacking encryption and authentication, which, combined with an expanding attack surface, necessitates robust security measures. In response to this challenge, numerous Intrusion Detection Systems (IDS) have been developed and deployed. Nonetheless, an open, comprehensive, and realistic dataset to test the effectiveness of such IDSs remains absent in the existing literature. This paper addresses this gap by considering the latest ROAD dataset, containing stealthy and sophisticated injections. The methodology involves dataset labelling and the implementation of both state-of-the-art deep learning models and traditional machine learning models to show the discrepancy in performance between the datasets most commonly used in the literature and the ROAD dataset, a more realistic alternative.

cs.CR↗

arXiv2024

Debiasing surgeon: fantastic weights and how to find them

Nowadays an ever-growing concerning phenomenon, the emergence of algorithmic biases that can lead to unfair models, emerges. Several debiasing approaches have been proposed in the realm of deep learning, employing more or less sophisticated approaches to discourage these models from massively employing these biases. However, a question emerges: is this extra complexity really necessary? Is a vanilla-trained model already embodying some ``unbiased sub-networks'' that can be used in isolation and propose a solution without relying on the algorithmic biases? In this work, we show that such a sub-network typically exists, and can be extracted from a vanilla-trained model without requiring additional training. We further validate that such specific architecture is incapable of learning a specific bias, suggesting that there are possible architectural countermeasures to the problem of biases in deep neural networks.

cs.LG↗

arXiv2023

SCoTTi: Save Computation at Training Time with an adaptive framework

On-device training is an emerging approach in machine learning where models are trained on edge devices, aiming to enhance privacy protection and real-time performance. However, edge devices typically possess restricted computational power and resources, making it challenging to perform computationally intensive model training tasks. Consequently, reducing resource consumption during training has become a pressing concern in this field. To this end, we propose SCoTTi (Save Computation at Training Time), an adaptive framework that addresses the aforementioned challenge. It leverages an optimizable threshold parameter to effectively reduce the number of neuron updates during training which corresponds to a decrease in memory and computation footprint. Our proposed approach demonstrates superior performance compared to the state-of-the-art methods regarding computational resource savings on various commonly employed benchmarks and popular architectures, including ResNets, MobileNet, and Swin-T.

cs.LG↗

arXiv2023

Enhanced EEG-Based Mental State Classification : A novel approach to eliminate data leakage and improve training optimization for Machine Learning

In this paper, we explore prior research and introduce a new methodology for classifying mental state levels based on EEG signals utilizing machine learning (ML). Our method proposes an optimized training method by introducing a validation set and a refined standardization process to rectify data leakage shortcomings observed in preceding studies. Furthermore, we establish novel benchmark figures for various models, including random forest and deep neural networks.

eess.SP↗