Graph pooling has gained significant progress in recent years as an effective solution for graph-level property classification tasks. With the emergence of research on Heterogeneous Information Networks (HINs), this paper argues that graph-level datasets for graph classification should be treated as HINs rather than homogeneous graphs to enhance information aggregation. We propose HINPool, a novel and general graph pooling framework for graph-level property classification with HINs. First, we devise a systematic HIN construction procedure from the original data to capture complex interactions. Next, we introduce a type-aware heterogeneous graph pooling method featuring a Type-Aware Selector (TAS) to select essential nodes and a Readout Aggregator (RA) to fuse critical information into a graph-level representation. Finally, a cross-layer fusion function is applied to combine the output embeddings from each graph pooling layer, creating a final graph representation for downstream classification tasks. Our approach achieves near state-of-the-art performance on widely used graph classification benchmark datasets, demonstrating significant improvements in four out of five datasets. This work redefines the strategy for graph-level property classification with HGNNs and heterogeneous graph pooling to model intricate relationships, enhancing performance without requiring extensive domain-specific knowledge.
High-performance and high-precision flow monitoring is a crucial function for network management, network bandwidth usage accounting and billing, network security, network forensics, and other important tasks. Nowadays, many commercial switches/routers provide either sFlow, NetFlow, or IPFIX scheme for monitoring the flows traversing a network. sFlow is a scheme widely supported by many switches/routers due to its using a sampling-based method, which greatly reduces the CPU processing load on a switch/router and the network bandwidth required to transmit flow data to a remote collector. However, many small flows may go undetected and the estimated flow data (e.g., the packet count and byte count) for detected flows can significantly deviate from their ground truth. NetFlow, which is Cisco Systems’ proprietary technology, does not use a sampling-based method by default. Instead, it tries to collect complete and correct flow data for every flow. However, as the link speed and the flow arrival rate continue to increase, NetFlow also provides a sampling-based option to reduce the CPU utilization of the switch/router. Because NetFlow is proprietary, an Internet Engineering Task Force (IETF) working group has defined IPFIX as an open flow information export protocol based on NetFlow Version 9. The requirements for IPFIX are defined in the RFC 3917 standards. Basically, IPFIX is the same as NetFlow Version 9. Due to its high demand on the CPU of the switch/router, currently NetFlow is supported only on very high-end switches/routers and its design and implementation on these commercial switches/routers are not published in the literature. In this paper, we design and implement a high-performance and high-precision NetFlow/IPFIX system on a Programming Protocol-independent Packet Processors (P4) hardware switch. Based on a 20 Gbps playback of a packet trace gathered on an Internet backbone link, experimental results show that our novel method significantly outperforms the typical design and implementation method of NetFlow/IPFIX on a P4 hardware switch. For example, for the number of detected flows during the trace period, our method outperforms the typical method by a factor of 5.72. As for the number of flows whose packet and byte counts are correctly counted, our method outperforms the typical method by a factor of 8.57.
Biometric recognition plays an increasingly pivotal role in cybersecurity, where the CIA triad, Confidentiality, Integrity, and Availability, forms the cornerstone of information security, with authentication as a critical yet challenging component. This paper presents the Biometric Multi-modal Authentication System using Geometric Programming (BMMA-GPT), tailored for deployment in Fast IDentity Online (FIDO/FIDO2)-enabled environments and Zero Trust Architectures (ZTA). The system employs a dual-threshold mechanism integrated with Defense-in-Depth (DiD) strategies to simultaneously enhance accuracy, efficiency, and security. The underlying optimization problem is formulated as a mathematical programming task and reformulated into a Geometric Programming (GP) model to efficiently compute optimal biometric permutations and verification thresholds under constrained estimation errors. BMMA-GPT enables the flexible integration of multiple biometric modalities, allowing dynamic adjustments to meet both individual user profiles and organizational security requirements. It achieves a high Area Under Curve (AUC) of approximately 0.99 while maintaining authentication latency under 1.5 seconds. This design supports Chief Information Security Officers (CISOs) in configuring tailored authentication processes with minimal computational cost, enhancing resilience against spoofing attacks and ensuring seamless user experience. By aligning biometric verification with DiD principles and GP-based optimization, the proposed framework offers a scalable and robust solution for identity authentication in complex digital ecosystems.
A multitude of interconnected risk events---ranging from regulatory changes to geopolitical tensions---can trigger ripple effects across firms. Identifying inter-firm risk relations is thus crucial for applications like portfolio management and investment strategy. Traditionally, such assessments rely on expert judgment and manual analysis, which are, however, subjective, labor-intensive, and difficult to scale. To address this, we propose a systematic method for extracting inter-firm risk relations using Form 10-K filings---authoritative, standardized financial documents---as our data source. Leveraging recent advances in natural language processing, our approach captures implicit and abstract risk connections through unsupervised fine-tuning based on chronological and lexical patterns in the filings. This enables the development of a domain-specific financial encoder with a deeper contextual understanding and introduces a quantitative risk relation score for transparency, interpretable analysis. Extensive experiments demonstrate that our method outperforms strong baselines across multiple evaluation settings.
The fusion of tiny energy harvesting devices with deep neural networks (DNN) optimized for intermittent execution is vital for sustainable intelligent applications at the edge. However, current intermittent-aware neural architecture search (NAS) frameworks overlook the inherent intermittency management overhead (IMO) of DNNs, leading to under-performance upon deployment. Moreover, we observe that straightforward IMO minimization within NAS may degrade solution accuracy. This work explores the relationship between DNN architectural characteristics, IMO, and accuracy, uncovering the varying sensitivity toward IMO across different DNN characteristics. Inspired by our insights, we present two guidelines for leveraging IMO sensitivity in NAS. First, the overall architecture search space can be reduced to exclude parameters with low IMO sensitivity, and second, network blocks with high IMO sensitivity can be primarily focused during the search, facilitating the discovery of highly accurate networks with low IMO. We incorporate these guidelines into TiNAS, which integrates cutting-edge tiny NAS and intermittent-aware NAS frameworks. Evaluations are conducted across various datasets and latency requirements, as well as deployment experiments on a Texas Instruments device under different intermittent power profiles. Compared to two variants, one minimizing IMO and the other disregarding IMO, TiNAS, respectively, achieves up to 38% higher accuracy and 33% lower IMO, with greater improvements for larger datasets. Its deployed solutions also achieve up to a 1.33 times inference speedup, especially under fluctuating power conditions.
In this paper, we examine the existence of the Rényi divergence between two time invariant hidden Markov models with arbitrary positive initial distributions. By making use of a Markov chain representation of the probability distribution for the hidden Markov model and eigenvalue for the associated Markovian operator, we obtain, under some regularity conditions, convergence of the Rényi divergence. By using this device, we also characterize the Rényi divergence and obtain the Kullback–Leibler divergence as of the Rényi divergence. Several examples, including classical finite state hidden Markov models, Markov switching models, and recurrent neural networks, are given for illustration. Moreover, we develop a non-Monte Carlo method that computes the Rényi divergence of two-state Markov switching models via the underlying invariant probability measure, which is characterized by the Fredholm integral equation.
Tiny battery-free devices running deep neural networks (DNNs) embody intermittent TinyML, a paradigm at the intersection of intermittent computing and deep learning, bringing sustainable intelligence to the extreme edge. This paper, as an overview of a special session at Embedded Systems Week (ESWEEK) 2025, presents four tales from diverse research backgrounds, sharing experiences in addressing unique challenges of efficient and reliable DNN inference despite the intermittent nature of ambient power. The first explores enhancing inference engines for efficient progress accumulation in hardware-accelerated intermittent inference and designing networks tailored for such execution. The second investigates computationally light, adaptive algorithms for faster, energy-efficient inference, and emerging computing-in-memory architectures for power failure resiliency. The third addresses battery-free networking, focusing on timely neighbor discovery and maintaining synchronization despite spatio-temporal energy dynamics across nodes. The fourth leverages modern nonvolatile memory fault behavior and DNN robustness to save energy without significant accuracy loss, with applicability to intermittent inference on nano-satellites. Collectively, these early efforts advance intermittent TinyML research and promote future cross-domain collaboration to tackle open challenges.
Guaranteeing reliable deep neural network (DNN) inference despite intermittent power is the cornerstone of enabling intelligent systems in energy-harvesting environments. Existing intermittent inference approaches support static neural networks with deterministic execution characteristics, accumulating progress across power cycles. However, dynamic neural networks adapt their structures at runtime. We observe that because intermittent inference approaches are unaware of this non-deterministic execution behavior, they suffer from incorrect progress recovery, degrading inference accuracy and performance. This work proposes non-deterministic inference progress accumulation to enable dynamic neural network inference on intermittent systems. Our middleware, NodPA, realizes this methodology by strategically selecting additional progress information to capture the non-determinism of the power-interrupted computation while preserving only the changed portions of the progress information to maintain low runtime overhead. Evaluations are conducted on a Texas Instruments device with both static and dynamic neural networks under time-varying power sources. Compared to intermittent inference approaches reliant on determinism, NodPA is less prone to inference non-termination and achieves an average inference speedup of 1.57 times without compromising accuracy, with greater improvements for highly dynamic networks under weaker power.
The prevalence of hearing aids is increasing. However, optimizing their amplification remains challenging due to the complexity of integrating multiple components in traditional methods. To address this, we present NeuroAMP, a novel deep neural network for end-to-end, personalized amplification in hearing aids. NeuroAMP leverages spectral features and the listener’s audiogram as inputs, and we explore four architectures: Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Convolutional Recurrent Neural Network (CRNN), and Transformer. We also introduce Denoising NeuroAMP, an extension that integrates noise reduction with amplification for improved real-world performance. To enhance generalization, we employed a comprehensive data augmentation strategy during training on diverse speech (TIMIT, TMHINT) and music (Cadenza Challenge MUSIC) datasets. Evaluation using the Hearing Aid Speech Perception Index (HASPI), Hearing Aid Speech Quality Index (HASQI), and Hearing Aid Audio Quality Index (HAAQI) shows that the Transformer-based NeuroAMP achieves the best performance, with SRCC scores of 0.9927 (HASQI) and 0.9905 (HASPI) on TIMIT, and 0.9738 (HAAQI) on Cadenza dataset. Notably, the augmentation strategy maintains robust performance on unseen datasets (e.g., VoiceBank-DEMAND, MUSDB18-HQ). Furthermore, Denoising NeuroAMP outperforms both the conventional NAL-R+WDRC method and a two-stage baseline on the VoiceBank-DEMAND dataset, achieving HASPI of 0.90 and HASQI of 0.59. These results highlight the strong potential of NeuroAMP and Denoising NeuroAMP to provide a novel and effective framework for personalized hearing aid amplification.
Tags play a critical role in enhancing product discoverability, optimizing search results, and enriching recommendation systems on e-commerce platforms. Despite the recent advancements in large language models (LLMs), which have shown proficiency in processing and understanding textual information, their application in tag generation remains an under-explored yet complex challenge. To this end, we introduce a novel method for automatic product tagging using LLMs to create behavior-enhanced tags (BETags). Specifically, our approach begins by generating base tags using an LLM. These base tags are then refined into BETags by incorporating user behavior data. This method aligns the tags with users' actual browsing and purchasing behavior, enhancing the accuracy and relevance of tags to user preferences. By personalizing the base tags with user behavior data, BETags are able to capture deeper behavioral insights, which is essential for understanding nuanced user interests and preferences in e-commerce environments. Moreover, since BETags are generated offline, they do not impose real-time computational overhead and can be seamlessly integrated into downstream tasks commonly associated with recommendation systems and search optimization. Our evaluation of BETag across three datasets--- Amazon (Scientific), MovieLens-1M, and FreshFood---shows that our approach significantly outperforms both human-annotated tags and other automated methods. These results highlight BETag as a scalable and efficient solution for personalized automated tagging, advancing e-commerce platforms by creating more tailored and engaging user experiences.