Papers

Precise control in modern robotic applications is always an open issue due to unknown time-varying disturbances. Existing meta-learning-based approaches require a shared representation of environmental structures, which lack flexibility for realistic non-structural disturbances. Besides, representation error and the distribution shifts can lead to heavy degradation in prediction accuracy. This work presents a generalizable disturbance estimation framework that builds on meta-learning and feedback-calibrated online adaptation. By extracting features from a finite time window of past observations, a unified representation that effectively captures general non-structural disturbances can be learned without predefined structural assumptions. The online adaptation process is subsequently calibrated by a state-feedback mechanism to attenuate the learning residual originating from the representation and generalizability limitations. Theoretical analysis shows that simultaneous convergence of both the online learning error and the disturbance estimation error can be achieved. Through the unified meta-representation, our framework effectively estimates multiple rapidly changing disturbances, as demonstrated by quadrotor flight experiments. See the project page for video, supplementary material and code: https://nonstructural-metalearn.github.io.

This paper discusses the task of face-based speech synthesis, a kind of personalized speech synthesis where the synthesized voices are constrained to perceptually match with a reference face image. Due to the lack of TTS-quality audio-visual corpora, previous approaches suffer from either low synthesis quality or domain mismatch induced by a knowledge transfer scheme. This paper proposes a new approach called Vclip that utilizes the facial-semantic knowledge of the CLIP encoder on noisy audio-visual data to learn the association between face and voice efficiently, achieving 89.63% cross-modal verification AUC score on Voxceleb testset. The proposed method then uses a retrieval-based strategy, combined with GMM-based speaker generation module for a downstream TTS system, to produce probable target speakers given reference images. Experimental results demonstrate that the proposed Vclip system in conjunction with the retrieval step can bridge the gap between face and voice features for face-based speech synthesis. And using the feedback information distilled from downstream TTS helps to synthesize voices that match closely with reference faces. Demos available at sos1sos2sixteen.github.io/vclip.

Real-world physics can only be analytically modeled with a certain level of precision for modern intricate robotic systems. As a result, tracking aggressive trajectories accurately could be challenging due to the existence of residual physics during controller synthesis. This paper presents a self-supervised residual learning and trajectory optimization framework to address the aforementioned challenges. At first, unknown dynamic effects on the closed-loop model are learned and treated as residuals of the nominal dynamics, jointly forming a hybrid model. We show that learning with analytic gradients can be achieved using only trajectory-level data while enjoying accurate long-horizon prediction with an arbitrary integration step size. Subsequently, a trajectory optimizer is developed to compute the optimal reference trajectory with the residual physics along it minimized. It ends up with trajectories that are friendly to the following control level. The agile flight of quadrotors illustrates that by utilizing the hybrid dynamics, the proposed optimizer outputs aggressive motions that can be precisely tracked.

AV2 is the successor to the AV1 video coding standard developed by the Alliance for Open Media (AOMedia). Its primary objective is to deliver substantial compression gains and subjective quality improvements while maintaining low-complexity encoder and decoder operations. This paper describes the transform, quantization and entropy coding design in AV2, including redesigned transform kernels and data-driven transforms, expanded transform partitioning, and a mode & coefficient dependent transform signaling. AV2 introduces several new coding tools including Intra/Inter Secondary Transforms (IST), Trellis Coded Quantization (TCQ), Adaptive Transform Coding (ATC), Probability Adaptation Rate Adjustment (PARA), Forward Skip Coding (FSC), Cross Chroma Component Transforms (CCTX), Parity Hiding (PH) tools and improved lossless coding. These advances enable AV2 to deliver the highest quality video experience for video applications at a significantly reduced bitrate.

Optimal power flow (OPF) is one of the fundamental tasks for power system operations. While machine learning (ML) approaches such as deep neural networks (DNNs) have been widely studied to enhance OPF solution speed and performance, their practical deployment faces two critical scaling questions: What is the minimum training data volume required for reliable results? How should ML models' complexity balance accuracy with real-time computational limits? Existing studies evaluate discrete scenarios without quantifying these scaling relationships, leading to trial-and-error-based ML development in real-world applications. This work presents the first systematic scaling study for ML-based OPF across two dimensions: data scale (0.1K-40K training samples) and compute scale (multiple NN architectures with varying FLOPs). Our results reveal consistent power-law relationships on both DNNs and physics-informed NNs (PINNs) between each resource dimension and three core performance metrics: prediction error (MAE), constraint violations and speed. We find that for ACOPF, the accuracy metric scales with dataset size and training compute. These scaling laws enable predictable and principled ML pipeline design for OPF. We further identify the divergence between prediction accuracy and constraint feasibility and characterize the compute-optimal frontier. This work provides quantitative guidance for ML-OPF design and deployments.

Smart grid infrastructure needs improved resilience and preventive maintenance through more accurate predictions. Current methodologies lack accurate representation of spatio-temporal-causal interdependencies and class imbalance in failure prediction tasks. This study introduces a Topology-Aware Spatio-Temporal Graph Transformer (ST-GT) architecture that overcomes existing limitations by using three main innovations: (1) directly incorporating physical transmission network topology into the transformer attention mechanism to identify spatial failure propagation patterns; (2) unified processing of static topological descriptors and (3) temporal Phasor Measurement Units (PMU) sequences in an end-to-end framework. The ST-GT model exhibited outstanding performance in five-fold cross-validation across 10 substations, attaining perfect recall (1.000 $\pm$ 0.001) and an F1-score of 0.858 $\pm$ 0.009, markedly surpassing XGBoost baselines (0.683 accuracy/F1). Perfect recall guarantees that no critical failures are overlooked, which is essential for grid safety; however, it may lead to an increase in false alarm rates. This framework integrates temporal dynamics modeling with spatial graph awareness for critical infrastructure monitoring. It offers interpretable insights into failure propagation pathways and enhances maintenance strategies. Future research focuses on developing cost-weighted loss functions for precision-recall trade-off enhancement, implementing real-time monitoring systems with uncertainty quantification, and creating cost-sensitive frameworks balancing false alarm expenditures with failure consequences. The methodology success suggests its potential for wider application in critical infrastructure areas requiring spatial temporal failure prediction.

Electroencephalography (EEG) foundation models hold significant promise for universal Brain-Computer Interfaces (BCIs). However, existing approaches often rely on end-to-end fine-tuning and exhibit limited efficacy under frozen-probing protocols, lacking the intrinsic universality required for broad generalization. This limitation stems from adapting general-purpose sequence architectures that overlook the biophysical and dynamical principles of neural activity. To bridge this gap, we propose DeeperBrain, a neuro-grounded foundation model integrating domain-specific inductive biases into its model design and learning objectives. Architecturally, DeeperBrain incorporates a volume conduction-aware channel encoding to model spatial mixing via 3D geometry, and a neurodynamics-aware temporal encoding capturing slow adaptations using oscillatory and exponential bases. For pretraining, we introduce a dual-objective strategy combining Masked EEG Reconstruction (MER) for local fidelity and Neurodynamics Statistics Prediction (NSP). NSP enforces alignment with macroscopic brain states by predicting interpretable order parameters, including spectral power, functional connectivity, cross-frequency coupling, and dynamic complexity. Extensive experiments demonstrate that DeeperBrain achieves state-of-the-art or highly competitive performance under end-to-end fine-tuning. Crucially, it maintains superior efficacy under a rigorous frozen-probing protocol, verifying that embedding neuroscientific first principles endows learned representations with the intrinsic universality essential for universal BCI. The code will be publicly available.

This paper deals with the gradient-based extremum seeking control (ESC) with actuation dynamics governed by distributed wave partial differential equations (PDEs). To achieve the control objective of real-time optimization for this class of infinite-dimensional systems, we first solve the trajectory generation problem to re-design the additive perturbation signal of the ESC system. Then, we develop a boundary control law through the backstepping method to compensate for the wave PDE with distributed effects, which ensures the exponential stability of the average closed-loop system by means of a Lyapunov-based analysis. At last, by employing the averaging theory for infinite-dimensional systems, we prove that the closed-loop trajectories converge to a small neighborhood surrounding the optimal point. Numerical simulations are presented to illustrate the effectiveness of the proposed method.

This paper presents a measurement-based framework for characterizing altitude-dependent spectral behavior of signals received by a tethered Helikite unmanned aerial vehicle (UAV). Using a multi-year spectrum measurement campaign in an outdoor urban environment, power spectral density snapshots are collected over the 89 MHz–6 GHz range. Three altitude-dependent spectral metrics are extracted: band-average power, spectral entropy, and spectral sparsity. We introduce the Altitude-Dependent Spectral Structure Model (ADSSM) to characterize the spectral power and entropy using first-order altitude-domain differential equations, and spectral sparsity using a logistic function, yielding closed-form expressions with physically consistent asymptotic behavior. The model is fitted to altitude-binned measurements from three annual campaigns at the AERPAW testbed across six licensed and unlicensed sub-6 GHz bands. Across all bands and years, the ADSSM achieves low root-mean-square error and high coefficients of determination. Results indicate that power transitions occur over narrow low-altitude regions, while entropy and sparsity evolve over broader, band-dependent altitude ranges, demonstrating that altitude-dependent spectrum behavior is inherently multidimensional. By explicitly modeling altitude-dependent transitions in spectral structure beyond received power, the proposed framework enables spectrum-aware UAV sensing and band selection decisions that are not achievable with conventional power- or threshold-based occupancy models.

Solving inverse problems in imaging requires models that support efficient inference, uncertainty quantification, and principled probabilistic reasoning. Energy-Based Models (EBMs), with their interpretable energy landscapes and compositional structure, are well-suited for this task but have historically suffered from high computational costs and training instability. To overcome the historical shortcomings of EBMs, we introduce a fast distillation strategy to transfer the strengths of pre-trained diffusion models into multi-scale EBMs. These distilled EBMs enable efficient sampling and preserve the interpretability and compositionality inherent to potential-based frameworks. Leveraging EBM compositionality, we propose Annealed Langevin Posterior Sampling (ALPS) algorithm for Maximum-A-Posteriori (MAP), Minimum Mean Square Error (MMSE), and uncertainty estimates for inverse problems in imaging. Unlike diffusion models that use complex guidance strategies for latent variables, we perform annealing on static posterior distributions that are well-defined and composable. Experiments on image inpainting and MRI reconstruction demonstrate that our method matches or surpasses diffusion-based baselines in both accuracy and efficiency, while also supporting MAP recovery. Overall, our framework offers a scalable and principled solution for inverse problems in imaging, with potential for practical deployment in scientific and clinical settings. ALPS code is available at the GitHub repository \href{https://github.com/JyoChand/ALPS}{ALPS}.

Artificial intelligence systems are increasingly embedded as persistent, closed-loop components within cyber-physical, social, and institutional processes. Rather than producing isolated outputs, such systems operate continuously under feedback, adaptation, and scale, reshaping physical flows, human behavior, and institutional practice over time. In these settings, socially unacceptable outcomes rarely arise from singular faults or explicit policy violations. Instead, they emerge through cumulative execution trajectories enabled by repetition, concurrency, and feedback. This paper advances the formal foundation of the Social Responsibility Stack (SRS) by making its central requirement explicit: responsibility is fundamentally a reachability property of system execution. A system is responsible iff its execution semantics prevent entry into inadmissible global configurations, regardless of local performance gains or optimization objectives. Responsibility failures are therefore not objective-level errors, but execution-level failures of trajectory control. To operationalize this perspective, we introduce Petri nets as an execution-level formalism for responsible autonomous systems. We show how SRS value commitments correspond to forbidden markings, safeguards to structural constraints on transition firing, auditing to monitoring of reachability pressure, and governance to legitimate modification of execution structure. Embedding Petri-net reachability within the SRS architecture internalizes responsibility as a structural invariant rather than an external objective or post-hoc mechanism. These results establish the Social Responsibility Stack as an executable responsibility architecture and position reachability-based execution semantics as a necessary foundation for responsible autonomy in feedback-rich cyber-physical and socio-technical systems.

The growing demand for mid-band spectrum necessitates efficient Dynamic Spectrum Sharing (DSS) to ensure coexistence between cellular networks and incumbent radar systems. Existing Spectrum Access System (SAS) frameworks rely on fixed Environmental Sensing Capability (ESC) sensors, which are latency-prone and inflexible. This paper introduces O-DSS, an O-RAN-compliant, Machine Learning (ML)-driven DSS framework that enables real-time cellular-radar coexistence in mid-band frequencies with shipborne and fast-moving airborne radars. O-DSS integrates radar detection from low-overhead Key Performance Metrics (KPMs) with spectrogram-based localization to drive fine-grained RAN control, including PRB blanking and radar-aware MCS adaptation. Deployed as a modular xApp, O-DSS achieves 60~ms detection and 700~ms evacuation latencies, outperforming existing baselines. Evaluations across simulations and Over-The-Air (OTA) testbeds show that O-DSS ensures robust incumbent protection while maintaining cellular performance by achieving radar detection of $\geq 99\%$ at SINR $\geq -4$~dB and localization recall of $\geq 95\%$ at SINR $\geq 8$~dB.

In this study, we evaluated four binarization methods. Locality-Sensitive Hashing (LSH), Iterative Quantization (ITQ), Kernel-based Supervised Hashing (KSH), and Supervised Discrete Hashing (SDH) on the ODIR dataset using deep feature embeddings. Experimental results show that SDH achieved the best performance, with an mAP@100 of 0.9184 using only 32-bit codes, outperforming LSH, ITQ, and KSH. Compared with prior studies, our method proved highly competitive: Fang et al. reported 0.7528 (Fundus-iSee, 48 bits) and 0.8856 (ASOCT-Cataract, 48 bits), while Wijesinghe et al. achieved 94.01 (KVASIR, 256 bits). Despite using significantly fewer bits, our SDH-based framework reached retrieval accuracy close to the state-of-the-art. These findings demonstrate that SDH is the most effective approach among those tested, offering a practical balance of accuracy, storage, and efficiency for medical image retrieval and device inventory management.

The rapid growth of dermatological imaging and mobile diagnostic tools calls for systems that not only demonstrate empirical performance but also provide strong theoretical guarantees. Deep learning models have shown high predictive accuracy; however, they are often criticized for lacking well, calibrated uncertainty estimates without which these models are hardly deployable in a clinical setting. To this end, we present the Conformal Bayesian Dermatological Classifier (CBDC), a well, founded framework that combines Statistical Learning Theory, Topological Data Analysis (TDA), and Bayesian Conformal Inference. CBDC offers distribution, dependent generalization bounds that reflect dermatological variability, proves a topological stability theorem that guarantees the invariance of convolutional neural network embeddings under photometric and morphological perturbations and provides finite conformal coverage guarantees for trustworthy uncertainty quantification. Through exhaustive experiments on the HAM10000, PH2, and ISIC 2020 datasets, we show that CBDC not only attains classification accuracy but also generates calibrated predictions that are interpretable from a clinical perspective. This research constitutes a theoretical and practical leap for deep dermatological diagnostics, thereby opening the machine learning theory clinical applicability interface.

This paper presents a new HPDOb that significantly improves disturbance estimation accuracy and robustness in motion control systems, surpassing the capabilities of conventional DObs. The proposed observer is analysed and synthesised in the discrete-time domain, providing a realistic representation of their dynamic behaviour and enabling enhanced controller design for practical applications. The core contribution of the HPDOb is a novel synthesis method that incorporates higher-order truncation error dynamics into disturbance estimation. Unlike conventional DObs, which are limited to zero-order truncation error, the HPDOb achieves first-order truncation error, yielding markedly improved estimation accuracy and robustness against disturbances in motion control systems. Simulation and experiments verify the stability and performance of HPDOb.

This paper presents robust position control strategies for the novel VSSEA. By employing a constructed state-space model, two control schemes are developed in a unified framework: a state-feedback controller and a sliding mode controller, both integrated with a second-order DOb. The proposed framework achieves high-performance motion control by precisely estimating and compensating for internal and external disturbances, while preserving the nominal dynamic response. Simulation results demonstrate that pole-placement-based controllers are highly sensitive to disturbances, whereas LQR-based controllers offer improved robustness at the expense of slower dynamics. By incorporating DOb, robustness is significantly enhanced without degrading time response, and the LQR controller can be tuned solely for performance optimization. Experimental results confirm that the proposed robust position controllers can be implemented in real world applications. These results highlight the effectiveness of the proposed approach and lay the foundation for future investigations on robust stability and performance under different stiffness settings.

Purpose: Annotation of medical breast images is an essential step toward better diagnostic but a time consuming task. This research aims to focus on different selecting sample strategies within deep active learning on Breast Region Segmentation (BRS) to lessen computational cost of training and effective use of resources. Methods: The Stavanger breast MRI dataset containing 59 patients was used in this study, with FCN-ResNet50 adopted as a sustainable deep learning (DL) model. A novel sample selection approach based on Breast Anatomy Geometry (BAG) analysis was introduced to group data with similar informative features for DL. Patient positioning and Breast Size were considered the key selection criteria in this process. Four selection strategies including Random Selection, Nearest Point, Breast Size, and a hybrid of all three strategies were evaluated using an active learning framework. Four training data proportions of 10%, 20%, 30%, and 40% were used for model training, with the remaining data reserved for testing. Model performance was assessed using Dice score, Intersection over Union, precision, and recall, along with 5-fold cross-validation to enhance generalizability. Results: Increasing the training data proportion from 10% to 40% improved segmentation performance for nearly all strategies, except for Random Selection. The Nearest Point strategy consistently achieved the lowest carbon footprint at 30% and 40% data proportions. Overall, combining the Nearest Point strategy with 30% of the training data provided the best balance between segmentation performance, efficiency, and environmental sustainability. Keywords: Deep Active Learning, Breast Region Segmentation, Human-center analysis

Layer-wise post-training quantization reconstructs each layer from inputs already altered by the quantized prefix. QEP compensates for this drift with one model-wide coefficient, conflating the model-level operating point with residual variation across layers. We present FADE, which constructs layer-specific coefficients from normalized round-to-nearest distortion and a heuristic calibrated-solver response. It requires no training or per-model coefficient search and adds no inference-time operation. We evaluate seven Whisper, Moonshine, and Qwen3-ASR models at 3 and 4 bits on four English ASR benchmarks. Across 38 settings, FADE lowers mean word error rate relative to fixed QEP-0.5 in 31, although a development-tuned global coefficient recovers much of this gap. The largest absolute reductions occur in 3-bit settings whose final error remains too high for practical use; these results measure collapse mitigation rather than deployable accuracy. In two lower-WER 3-bit cases, FADE reduces the tuned-global result from 3.67 to 3.10 and from 13.03 to 11.63. Most 4-bit differences from the tuned control are within a descriptive tolerance, and one reverses. Paired reruns and within-seed permutations on an outcome-informed subset support assignment sensitivity in selected cases, but do not estimate a matrix-wide success rate. FADE is therefore a layer-wise alternative when per-model coefficient search is unavailable, not a universal replacement for tuned global compensation.

The rapid advancement of speech synthesis technologies, including text-to-speech (TTS) and voice conversion (VC), has intensified security and privacy concerns related to voice cloning. Recent defenses attempt to prevent unauthorized cloning by embedding protective perturbations into speech to obscure speaker identity while maintaining intelligibility. However, adversaries can apply advanced purification techniques to remove these perturbations, recover authentic acoustic characteristics, and regenerate cloneable voices. Despite the growing realism of such attacks, the robustness of existing defenses under adaptive purification remains insufficiently studied. Most existing purification methods are designed to counter adversarial noise in automatic speech recognition (ASR) systems rather than speaker verification or voice cloning pipelines. As a result, they fail to suppress the fine-grained acoustic cues that define speaker identity and are often ineffective against speaker verification attacks (SVA). To address these limitations, we propose Diffusion-Bridge (VocalBridge), a purification framework that learns a latent mapping from perturbed to clean speech in the EnCodec latent space. Using a time-conditioned 1D U-Net with a cosine noise schedule, the model enables efficient, transcript-free purification while preserving speaker-discriminative structure. We further introduce a Whisper-guided phoneme variant that incorporates lightweight temporal guidance without requiring ground-truth transcripts. Experimental results show that our approach consistently outperforms existing purification methods in recovering cloneable voices from protected speech. Our findings demonstrate the fragility of current perturbation-based defenses and highlight the need for more robust protection mechanisms against evolving voice-cloning and speaker verification threats.

Multimodal large language models (MLLMs) show promising performance on medical visual question answering (VQA) and report generation, but these generation and explanation abilities do not reliably transfer to disease-specific classification. We evaluated MLLM architectures on knee osteoarthritis (OA) radiograph classification, which remains underrepresented in existing medical MLLM benchmarks, even though knee OA affects an estimated 300 to 400 million people worldwide. Through systematic ablation studies manipulating the vision encoder, the connector, and the large language model (LLM) across diverse training strategies, we measured each component's contribution to diagnostic accuracy. In our classification task, a trained vision encoder alone could outperform full MLLM pipelines in classification accuracy and fine-tuning the LLM provided no meaningful improvement over prompt-based guidance. And LoRA fine-tuning on a small, class-balanced dataset (500 images) gave better results than training on a much larger but class-imbalanced set (5,778 images), indicating that data balance and quality can matter more than raw scale for this task. These findings suggest that for domain-specific medical classification, LLMs are more effective as interpreters and report generators rather than as primary classifiers. Therefore, the MLLM architecture appears less suitable for medical image diagnostic classification tasks that demand high certainty. We recommend prioritizing vision encoder optimization and careful dataset curation when developing clinically applicable systems.

Background: Deep learning superresolution (SR) may enhance musculoskeletal MR image quality, but its diagnostic value in knee imaging at 7T is unclear. Objectives: To compare image quality and diagnostic performance of SR, low-resolution (LR), and high-resolution (HR) 7T knee MRI. Methods: In this prospective study, 42 participants underwent 7T knee MRI with LR (0.8*0.8*2 mm3) and HR (0.4*0.4*2 mm3) sequences. SR images were generated from LR data using a Hybrid Attention Transformer model. Three radiologists assessed image quality, anatomic conspicuity, and detection of knee pathologies. Arthroscopy served as reference in 10 cases. Results: SR images showed higher overall quality than LR (median score 5 vs 4, P<.001) and lower noise than HR (5 vs 4, P<.001). Visibility of cartilage, menisci, and ligaments was superior in SR and HR compared to LR (P<.001). Detection rates and diagnostic performance (sensitivity, specificity, AUC) for intra-articular pathology were similar across image types (P>=.095). Conclusions: Deep learning superresolution improved subjective image quality in 7T knee MRI but did not increase diagnostic accuracy compared with standard LR imaging.

Speech-based machine learning systems are sensitive to noise, complicating reliable deployment in emotion recognition and voice pathology detection. We evaluate the robustness of a hybrid quantum machine learning model, quanvolutional neural networks (QNNs) against classical convolutional neural networks (CNNs) under four acoustic corruptions (Gaussian noise, pitch shift, temporal shift, and speed variation) in a clean-train/corrupted-test regime. Using AVFAD (voice pathology) and TESS (speech emotion), we compare three QNN models (Random, Basic, Strongly) to a simple CNN baseline (CNN-Base), ResNet-18 and VGG-16 using accuracy and corruption metrics (CE, mCE, RCE, RmCE), and analyze architectural factors (circuit complexity or depth, convergence) alongside per-emotion robustness. QNNs generally outperform the CNN-Base under pitch shift, temporal shift, and speed variation (up to 22% lower CE/RCE at severe temporal shift), while the CNN-Base remains more resilient to Gaussian noise. Among quantum circuits, QNN-Basic achieves the best overall robustness on AVFAD, and QNN-Random performs strongest on TESS. Emotion-wise, fear is most robust (80-90% accuracy under severe corruptions), neutral can collapse under strong Gaussian noise (5.5% accuracy), and happy is most vulnerable to pitch, temporal, and speed distortions. QNNs also converge up to six times faster than the CNN-Base. To our knowledge, this is a systematic study of QNN robustness for speech under common non-adversarial acoustic corruptions, indicating that shallow entangling quantum front-ends can improve noise resilience while sensitivity to additive noise remains a challenge.

In music creation, rapid prototyping is essential for exploring and refining ideas, yet existing generative tools often fall short when users require both structural control and stylistic flexibility. Prior approaches in stem-to-stem generation can condition on other musical stems but offer limited control over rhythm, and timbre-transfer methods allow users to specify specific rhythms, but cannot condition on musical context. We introduce DARC, a generative drum accompaniment model that conditions both on musical context from other stems and explicit rhythm prompts such as beatboxing or tapping tracks. Using parameter-efficient fine-tuning, we augment STAGE, a state-of-the-art drum stem generator, with fine-grained rhythm control while maintaining musical context awareness.

In Large Language Models (LLMs), the number of parameters has grown exponentially in the past few years, e.g., from 1.5 billion parameters in GPT-2 to 175 billion in GPT-3 to possibly more than trillion in higher versions. This raises a significant challenge for implementation, especially for Edge devices. Unlike cloud computing, memory and processing power for Edge devices are very limited, which necessitates developing novel ideas to make such applications feasible. In this work, we investigate compressing weights with a special quantization that limits numbers to only power-of-two (PoT). This helps save a huge amount of memory as only exponents need to be stored, more importantly, it significantly reduces processing power by replacing costly multiplication with low cost bit shifting. To overcome performance loss due to this strict quantization, we investigate Quantization Aware Training (QAT) to enhance performance through additional training. Results on GPT-2 124M show a major enhancement for quantized PoT model after additional training, with a perplexity enhancement of 66% and BERT-Score loss to baseline GPT-2 of 1%. The memory saving is estimated to be 87.5% while the inference speed is expected to be 3-10x faster with PoT quantization versus full-precision.

Three-phase asynchronous motor are fundamental components in industrial systems, and their failure can lead to significant operational downtime and economic losses. Vibration and current signals are effective indicators for monitoring motor health and diagnosing faults. However, motors in real applications often operate under variable conditions such as fluctuating speeds and loads, which complicate the fault diagnosis process. This paper presents a comprehensive dataset collected from a three-phase asynchronous motor under various fault types and severities, operating under diverse speed and load conditions. The dataset includes both single faults and mechanical-electrical compound faults, such as rotor unbalance, stator winding short circuits, bearing faults, and their combinations. Data were acquired under both steady and transitional conditions, with signals including triaxial vibration, three-phase currents, torque, and key-phase signals. This dataset supports the development and validation of robust fault diagnosis methods for electric motors under realistic operating conditions.

Effective data center cooling is crucial for reliable operation; however, cooling systems often exhibit inefficiencies that result in excessive energy consumption. This paper presents a three-stage, physics-guided machine learning framework for identifying and reducing cooling energy waste in high-performance computing facilities. Using one year of 10-minute resolution operational data from the Frontier exascale supercomputer, we first train a monotonicity-constrained gradient boosting surrogate that predicts facility accessory power from coolant flow rates, temperatures, and server power. The surrogate achieves a mean absolute error of 0.026 MW and predicts power usage effectiveness within 0.01 of measured values for 98.7% of test samples. In the second stage, the surrogate serves as a physics-consistent baseline to quantify excess cooling energy, revealing approximately 85 MWh of annual inefficiency concentrated in specific months, hours, and operating regimes. The third stage evaluates guardrail-constrained counterfactual adjustments to supply temperature and subloop flows, demonstrating that up to 96% of identified excess can be recovered through small, safe setpoint changes while respecting thermal limits and operational constraints. The framework yields interpretable recommendations, supports counterfactual analyses such as flow reduction during low-load periods and redistribution of thermal duty across cooling loops, and provides a practical pathway toward quantifiable reductions in accessory power. The developed framework is readily compatible with model predictive control and provides a template that, with site-specific recalibration, could be adapted to other liquid-cooled data centers with different configurations and cooling requirements.

We derive a state-space characterization of all dynamic state-feedback controllers that make an equilibrium of a nonlinear input-affine continuous-time system locally exponentially stable. Specifically, any controller obtained as the sum of a linear state-feedback $u=Kx$, with $K$ stabilizing the linearized system, and the output of internal locally exponentially stable controller dynamics is itself locally exponentially stabilizing. Conversely, every dynamic state-feedback controller that locally exponentially stabilizes the equilibrium admits such a decomposition. The result can be viewed as a state-space nonlinear Youla-type parametrization specialized to local, rather than global, and exponential, rather than asymptotic, closed-loop stability. The residual locally exponentially stable controller dynamics can be implemented with stable recurrent neural networks and trained as neural ODEs to achieve high closed-loop performance in nonlinear control tasks.

We develop a mathematical framework for the optimal scheduling of flexible water desalination plants (WDPs) as hybrid generator-load resources. WDPs integrate thermal generation, membrane-based controllable loads, and renewable energy sources, offering unique operational flexibility for power system operations. They can simultaneously participate in two markets: selling desalinated water to a water utility, and bidirectionally transacting electricity with the grid based on their net electricity demand. We formulate the scheduling decision problem of a profit-maximizing WDP, capturing operational, technological, and market-based coupling between water and electricity flows. The threshold-based structure we derive provides computationally tractable coordination suitable for large-scale deployment, offering operational and economical insights into how thermal and membrane-based desalination colocated with renewables complementarily provide continuous bidirectional flexibility. The thresholds are analytically characterized in near closed form as explicit functions of technology and tariff parameters. We examine how small changes in the exogenous tariff and technology parameters affect the WDP's profit. Extensive simulations illustrate the optimal WDP's operation, profit, and water-electricity exchange, demonstrating significant improvements relative to benchmark algorithms.

Recent advances in speaker diarization exploit large pretrained foundation models, such as WavLM, to achieve state-of-the-art performance on multiple datasets. Systems like DiariZen leverage these rich single-channel representations, but are limited to single-channel audio, preventing the use of spatial cues available in multi-channel recordings. This work analyzes the impact of incorporating spatial information into a state-of-the-art single-channel diarization system by evaluating several strategies for conditioning the model on multi-channel spatial features. Experiments on meeting-style datasets indicate that spatial information can improve diarization performance, but the overall improvement is smaller than expected for the proposed system, suggesting that the features aggregated over all WavLM layers already capture much of the information needed for accurate speaker discrimination, also in overlapping speech regions. These findings provide insight into the potential and limitations of using spatial cues to enhance foundation model-based diarization.

This paper presents a novel channel-estimation (CE) method that mitigates residual phase drifts in backscatter links and a full hardware and signal-processing pipeline for a single-antenna monostatic system. The platform comprises a semi-passive tag, a software-defined radio (SDR) reader, and a 2x1 planar Yagi-Uda array (7 dBi with higher than 30 dB isolation) operating at 2.4 ~ 2.5 GHz. The developed backscatter fading model accounts for round-trip propagation and temporal correlation, and employs an analytically derived resource-optimal pilot allocation strategy. At the receiver, optimized least square (LS) and linear minimum mean square error (LMMSE) CE with pilot-aided carrier frequency offset (CFO) compensation feed a zero-forcing (ZF) equalizer to suppress ISI. The prototype delivers 500 kbps at 1 m with power of 158 uW (SDR baseband) and 10 uW (RF switch), yielding 320 pJ/bit. OOK and BPSK modulations achieve measured EVMs of 2.97 % and 4.02 %, respectively. Performance is validated by BER measurements and successful reconstruction of a full-color image in an over-the-air experiment. The results demonstrate an ultra-low-power, multimedia-capable backscatter IoT link and provide practical hardware-software co-design guidance for scalable deployments.