Search

This paper introduces StyleSpeech, a novel Text-to-Speech~(TTS) system that enhances the naturalness and accuracy of synthesized speech. Building upon existing TTS technologies, StyleSpeech incorporates a unique Style Decorator structure that enables deep learning models to simultaneously learn style and phoneme features, improving adaptability and efficiency through the principles of Lower Rank Adaptation~(LoRA). LoRA allows efficient adaptation of style features in pre-trained models. Additionally, we introduce a novel automatic evaluation metric, the LLM-Guided Mean Opinion Score (LLM-MOS), which employs large language models to offer an objective and robust protocol for automatically assessing TTS system performance. Extensive testing on benchmark datasets shows that our approach markedly outperforms existing state-of-the-art baseline methods in producing natural, accurate, and high-quality speech. These advancements not only pushes the boundaries of current TTS system capabilities, but also facilitate the application of TTS system in more dynamic and specialized, such as interactive virtual assistants, adaptive audiobooks, and customized voice for gaming. Speech samples can be found in https://style-speech.vercel.app

Vision-based Interfaces (VIs) are pivotal in advancing Human-Computer Interaction (HCI), particularly in enhancing context awareness. However, there are significant opportunities for these interfaces due to rapid advancements in multimodal Artificial Intelligence (AI), which promise a future of tight coupling between humans and intelligent systems. AI-driven VIs, when integrated with other modalities, offer a robust solution for effectively capturing and interpreting user intentions and complex environmental information, thereby facilitating seamless and efficient interactions. This PhD study explores three application cases of multimodal interfaces to augment context awareness, respectively focusing on three dimensions of visual modality: scale, depth, and time: a fine-grained analysis of physical surfaces via microscopic image, precise projection of the real world using depth data, and rendering haptic feedback from video background in virtual environments., Comment: 3 pages, MOBILEHCI Adjunct '24 26th International Conference on Mobile Human-Computer Interaction, September 30-October 3, 2024, Melbourne, VIC, Australia

Integrating physiological signals such as electroencephalogram (EEG), with other data such as interview audio, may offer valuable multimodal insights into psychological states or neurological disorders. Recent advancements with Large Language Models (LLMs) position them as prospective ``health agents'' for mental health assessment. However, current research predominantly focus on single data modalities, presenting an opportunity to advance understanding through multimodal data. Our study aims to advance this approach by investigating multimodal data using LLMs for mental health assessment, specifically through zero-shot and few-shot prompting. Three datasets are adopted for depression and emotion classifications incorporating EEG, facial expressions, and audio (text). The results indicate that multimodal information confers substantial advantages over single modality approaches in mental health assessment. Notably, integrating EEG alongside commonly used LLM modalities such as audio and images demonstrates promising potential. Moreover, our findings reveal that 1-shot learning offers greater benefits compared to zero-shot learning methods., Comment: 6 pages; UbiComp Companion '24, Companion of the 2024 ACM International Joint Conference on Pervasive and Ubiquitous Computing, October 5--9, 2024}{Melbourne, VIC, Australia

Surface sensing is widely employed in health diagnostics, manufacturing and safety monitoring. Advances in mobile sensing affords this potential for context awareness in mobile computing, typically with a single sensing modality. Emerging multimodal large-scale language models offer new opportunities. We propose MultiSurf-GPT, which utilizes the advanced capabilities of GPT-4o to process and interpret diverse modalities (radar, microscope and multispectral data) uniformly based on prompting strategies (zero-shot and few-shot prompting). We preliminarily validated our framework by using MultiSurf-GPT to identify low-level information, and to infer high-level context-aware analytics, demonstrating the capability of augmenting context-aware insights. This framework shows promise as a tool to expedite the development of more complex context-aware applications in the future, providing a faster, more cost-effective, and integrated solution., Comment: 6 pages. MOBILEHCI Adjunct '24, 26th International Conference on Mobile Human-Computer Interaction, September 30-October 3, 2024, Melbourne, VIC, Australia

A service-level objective (SLO) is a target performance metric of service that cloud vendors aim to ensure. Delivering optimized SLOs can enhance user satisfaction and improve the competitiveness of cloud vendors. As large language models (LLMs) are gaining increasing popularity across various fields, it is of great significance to optimize SLOs for LLM inference services. In this paper, we observe that adjusting the parameters of LLM inference engines can improve service performance, and the optimal parameter configurations of different services are different. Therefore, we propose SCOOT, an automatic performance tuning system to optimize SLOs for each LLM inference service by tuning the parameters of the inference engine. We first propose a generalized formulation of the tuning problem to handle various objectives and constraints between parameters, and SCOOT exploits the Bayesian optimization (BO) technique to resolve the problem via exploration and exploitation. Moreover, SCOOT adopts a random forest to learn hidden constraints during the tuning process to mitigate invalid exploration. To improve the tuning efficiency, SCOOT utilizes the parallel suggestion to accelerate the tuning process. Extensive experiments demonstrate that SCOOT can significantly outperform existing tuning techniques in SLO optimization while greatly improving the tuning efficiency.

The primary focus of this research is the discreet and subtle everyday contact interactions between mobile phones and their surrounding surfaces. Such interactions are anticipated to facilitate mobile context awareness, encompassing aspects such as dispensing medication updates, intelligently switching modes (e.g., silent mode), or initiating commands (e.g., deactivating an alarm). We introduce MicroCam, a contact-based sensing system that employs smartphone IMU data to detect the routine state of phone placement and utilizes a built-in microscope camera to capture intricate surface details. In particular, a natural dataset is collected to acquire authentic surface textures in situ for training and testing. Moreover, we optimize the deep neural network component of the algorithm, based on continual learning, to accurately discriminate between object categories (e.g., tables) and material constituents (e.g., wood). Experimental results highlight the superior accuracy, robustness and generalization of the proposed method. Lastly, we conducted a comprehensive discussion centered on our prototype, encompassing topics such as system performance and potential applications and scenarios., Comment: 28 pages

High-redshift blazars are the most powerful extragalactic astrophysical sources ever detected in the high-energy gamma-ray band. In this study, we present a temporal and spectral analysis of the high-redshift blazar B3 1343+451 based on 14 years of Fermi-LAT observations, spanning from 2008 August 4 to 2022 June 6 (MJD 54686-59733). We extract a seven-day binned $\gamma$-ray light curve in the energy range 0.1--500 GeV and identify seven outburst periods with a peak flux of $>4.32\times10^{-7} \rm ph \cdot cm^{-2} \cdot s^{-1}$. The highest seven day flux (above 100 MeV) reaches $(8.06\pm0.56)\times10^{-7} \rm erg \ cm^{-2} \ s^{-1}$ on MJD = 56,177.16, which is 10 times higher than the flux in the quiescent period. To understand the properties of distant blazar jets, we employ a standard one-zone leptonic scenario and model the multiwavelength spectral energy distributions of one quiescent and seven flaring periods. We find that the $\gamma$-ray spectrum is better reproduced if we assume that the dissipation region of the jet, $R_{\rm diss}$, is located within the molecular torus, where infrared emission is the dominant external photon field. We infer that the jets in higher-redshift blazars have larger power and kinetic energy, where the kinetic energy is significantly greater than the radiation power, and the jet production efficiency suggests that we need to lower the accretion efficiency. These results imply that B3 1343+451 may have a standard thin disk surrounding its massive black hole, and the jets of B3 1343+451 may not be fully explained by the Blandford--Payne process., Comment: 13 pages, 5 figures.Accepted for publication in APJ

We present the timing analysis of 10 archived \XMM observations with an exposure of $>40$ ks of Markarian 421. Mrk 421 is the brightest high-frequency-peaked BL Lac object (HBL) emitting in X-rays produced by electrons accelerated in the innermost regions of a relativistic jet pointing toward us. For each observation, we construct averaged X-ray spectra in 0.5--10 keV band, as well as 100 s binned light curves (LCs) in various subbands. During these observations, the source exhibited various intensity states differing by close to an order of magnitude in flux, with the fractional variability amplitude increasing with energy through the X-ray band. Bayesian power spectral density analysis reveals that the X-ray variability can be characterized by a colored noise, with an index ranging from $\sim-1.9$ to $-3.0$. Moreover, both the standard cross-correlation function and cross-spectral methods indicate that the amount of time lags increases with the energy difference between two compared LCs. A time-dependent two-zone jet model is developed to extract physical information from the X-ray emission of Mrk 421. In the model, we assume that the jet emission mostly comprises a quasi-stationary component and a highly variable one. Our results show that the two-zone model can simultaneously provide a satisfactory description for both the X-ray spectra and time lags observed in different epochs, with the model parameters constrained in a fully acceptable interval. We suggest that shocks within the jets may be the primary energy dissipation process responsible for triggering the rapid variability, although magnetic reconnection cannot be excluded., Comment: 33 pages, 12 figures, 6 tables; Accepted for publication in ApJ supplement series

Large language models (LLMs) iteratively generate text token by token, with memory usage increasing with the length of generated token sequences. The unpredictability of generation lengths makes it difficult to estimate the time and memory needed to process requests, posing a challenge for effective request scheduling. Conventional sequence-level scheduling (SLS) serves requests in a first-come first-served (FCFS) manner with static batching where requests with short generation lengths are delayed until those with long ones have finished generation, which hurts computational efficiency. Besides, to avoid out-of-memory (OOM) errors, SLS batches requests with a small batch size, which limits throughput. Recently proposed iteration-level scheduling (ILS) enhances computational efficiency with continuous batching to return completed requests timely and dynamically add new requests for processing. However, many ILS schedulers limit the number of parallel-processing requests to avoid OOM errors while achieving a fast inference speed, which compromises throughput. Moreover, existing SLS and ILS schedulers fail to balance the workload across multiple deployed LLM instances. To tackle these challenges, we propose slice-level scheduling (SCLS). By splitting the predefined maximal generation length limit into slices and serving batches slice by slice, it provides a precise range of serving time and memory usage for batched requests, laying the foundation for effective scheduling. Experiments confirm that compared with SLS and ILS schedulers, SCLS can improve throughput by up to 315.8% and greatly mitigate load imbalance with proposed batching and offloading algorithms., Comment: 13 pages, 22 figures

Nowadays, large language models (LLMs) are published as a service and can be accessed by various applications via APIs, also known as language-model-as-a-service (LMaaS). Without knowing the generation length of requests, existing serving systems serve requests in a first-come, first-served (FCFS) manner with a fixed batch size, which leads to two problems that affect batch serving efficiency. First, the generation lengths of requests in a batch vary, and requests with short generation lengths must wait for requests with long generation lengths to finish during the batch serving procedure. Second, requests with longer generation lengths consume more memory during serving. Without knowing the generation lengths of batched requests, the batch size is always set small to avoid the out-of-memory (OOM) error, thus preventing the GPU from being fully utilized. In this paper, we find that a significant number of popular applications in the LMaaS scenario have a positive correlation between the generation length and the length of raw user input. Based on this observation, we propose Magnus, which can accurately predict the request generation length with the user input length, application-level, and user-level semantic features. Accordingly, Magnus can achieve high request throughput by batching requests of similar generation lengths together with adaptive batch sizes. Besides, Magnus can also schedule batches with the highest response ratio next (HRRN) policy to reduce request response time. Experiments conducted on our testbed show that Magnus improves request throughput by up to 234\% and reduces response time by up to 89.7\% compared to baselines., Comment: 12 pages, 14 figures

Existing fuzzy extractors and similar methods provide an effective way for extracting a secret key from a user's biometric data, but are susceptible to impersonation attack: once a valid biometric sample is captured, the scheme is no longer secure. We propose a novel multi-factor fuzzy extractor that integrates both a user's secret (e.g., a password) and a user's biometrics in the generation and reconstruction process of a cryptographic key. We then employ this multi-factor fuzzy extractor to construct personal identity credentials which can be used in a new multi-factor authenticated key exchange protocol that possesses multiple important features. First, the protocol provides mutual authentication. Second, the user and service provider can authenticate each other without the involvement of the identity authority. Third, the protocol can prevent user impersonation from a compromised identity authority. Finally, even when both a biometric sample and the secret are captured, the user can re-register to create a new credential using a new secret (reusable/reissued identity credentials). Most existing works on multi-factor authenticated key exchange only have a subset of these features. We formally prove that the proposed protocol is semantically secure. Our experiments carried out on the finger vein dataset SDUMLA achieved a low equal error rate (EER) of 0.04%, a reasonable averaged computation time of 0.93 seconds for the user and service provider to authenticate and establish a shared session key, and a small communication overhead of only 448 bytes., Comment: 17 pages

Vector data management systems (VDMSs) have become an indispensable cornerstone in large-scale information retrieval and machine learning systems like large language models. To enhance the efficiency and flexibility of similarity search, VDMS exposes many tunable index parameters and system parameters for users to specify. However, due to the inherent characteristics of VDMS, automatic performance tuning for VDMS faces several critical challenges, which cannot be well addressed by the existing auto-tuning methods. In this paper, we introduce VDTuner, a learning-based automatic performance tuning framework for VDMS, leveraging multi-objective Bayesian optimization. VDTuner overcomes the challenges associated with VDMS by efficiently exploring a complex multi-dimensional parameter space without requiring any prior knowledge. Moreover, it is able to achieve a good balance between search speed and recall rate, delivering an optimal configuration. Extensive evaluations demonstrate that VDTuner can markedly improve VDMS performance (14.12% in search speed and 186.38% in recall rate) compared with default setting, and is more efficient compared with state-of-the-art baselines (up to 3.57 times faster in terms of tuning time). In addition, VDTuner is scalable to specific user preference and cost-aware optimization objective. VDTuner is available online at https://github.com/tiannuo-yang/VDTuner., Comment: Accepted by ICDE 2024

Recent methods utilize graph contrastive Learning within graph-structured user-item interaction data for collaborative filtering and have demonstrated their efficacy in recommendation tasks. However, they ignore that the difference relation density of nodes between the user- and item-side causes the adaptability of graphs on bilateral nodes to be different after multi-hop graph interaction calculation, which limits existing models to achieve ideal results. To solve this issue, we propose a novel framework for recommendation tasks called Bilateral Unsymmetrical Graph Contrastive Learning (BusGCL) that consider the bilateral unsymmetry on user-item node relation density for sliced user and item graph reasoning better with bilateral slicing contrastive training. Especially, taking into account the aggregation ability of hypergraph-based graph convolutional network (GCN) in digging implicit similarities is more suitable for user nodes, embeddings generated from three different modules: hypergraph-based GCN, GCN and perturbed GCN, are sliced into two subviews by the user- and item-side respectively, and selectively combined into subview pairs bilaterally based on the characteristics of inter-node relation structure. Furthermore, to align the distribution of user and item embeddings after aggregation, a dispersing loss is leveraged to adjust the mutual distance between all embeddings for maintaining learning ability. Comprehensive experiments on two public datasets have proved the superiority of BusGCL in comparison to various recommendation methods. Other models can simply utilize our bilateral slicing contrastive learning to enhance recommending performance without incurring extra expenses.

We obtain the entropy and the surface entropy of the axial products on $\mathbb{N}^d$ and the $d$-tree $T^d$ of two types of systems: the subshift and the multiplicative subshift.

More than six million people died of the COVID-19 by April 2022. The heavy casualties have put people on great and urgent alert and people try to find all kinds of information to keep them from being inflected by the coronavirus. This research tries to find out whether the mobile health text information sent to peoples devices is correct as smartphones becoming the major information source for people. The proposed method uses various mobile information retrieval and data mining technologies including lexical analysis, stopword elimination, stemming, and decision trees to classify the mobile health text information to one of the following classes: (i) true, (ii) fake, (iii) misinformative, (iv) disinformative, and (v) neutral. Experiment results show the accuracy of the proposed method is above the threshold value 50 percentage, but is not optimal. It is because the problem, mobile text misinformation identification, is intrinsically difficult.

We calculate the Minkowski and Hausdorff dimensions of affine multiplicative subshifts on $\mathbb{N}$.

Recent years have seen the explosion of edge intelligence with powerful Deep Neural Networks (DNNs). One popular scheme is training DNNs on powerful cloud servers and subsequently porting them to mobile devices after being lightweight. Conventional approaches manually specialized DNNs for various edge platforms and retrain them with real-world data. However, as the number of platforms increases, these approaches become labour-intensive and computationally prohibitive. Additionally, real-world data tends to be sparse-label, further increasing the difficulty of lightweight models. In this paper, we propose MatchNAS, a novel scheme for porting DNNs to mobile devices. Specifically, we simultaneously optimise a large network family using both labelled and unlabelled data and then automatically search for tailored networks for different hardware platforms. MatchNAS acts as an intermediary that bridges the gap between cloud-based DNNs and edge-based DNNs.

Future food security is a major concern of the 21st century with the growing global population and climate changes. In addressing these challenges, protected cropping ensures food production year-round and increases crop production per land area by controlling environment conditions. Maintaining the growth and health of crops in these facilities is essential to ensure optimum food production. However, this is a laborious work and is currently done manually. Image-based non-destructive plant phenotyping is an emerging research area that reduces the skilled labour cost while enhancing the monitoring of crop growth, health, and identifying phenotype-genotype relations for plant breeding. With the proliferations of protected infrastructures and targeted plants, different technologies and sensor setups are needed for image-based crop monitoring. Conveyor-type plant-to-sensor systems, bench-top or gantry-based systems are commonly found in research facilities focussing on phenotyping of small, relatively short, or movable model plants. This review examines the literature on crop monitoring and phenotyping platforms in both field and protected facilities and explains different camera technologies and their ability to extract different plant traits. The review highlights the future research directions of image-based monitoring of commercial scale protected crops where crops can be relatively tall or vertically supported under semi controlled environments, which presents new challenges and is rarely covered in the literature.

Code Large Language Models (Code LLMs) have gained significant attention in the industry due to their wide applications in the full lifecycle of software engineering. However, the effectiveness of existing models in understanding non-English inputs for multi-lingual code-related tasks is still far from well studied. This paper introduces CodeFuse-13B, an open-sourced pre-trained code LLM. It is specifically designed for code-related tasks with both English and Chinese prompts and supports over 40 programming languages. CodeFuse achieves its effectiveness by utilizing a high quality pre-training dataset that is carefully filtered by program analyzers and optimized during the training process. Extensive experiments are conducted using real-world usage scenarios, the industry-standard benchmark HumanEval-x, and the specially designed CodeFuseEval for Chinese prompts. To assess the effectiveness of CodeFuse, we actively collected valuable human feedback from the AntGroup's software development process where CodeFuse has been successfully deployed. The results demonstrate that CodeFuse-13B achieves a HumanEval pass@1 score of 37.10%, positioning it as one of the top multi-lingual code LLMs with similar parameter sizes. In practical scenarios, such as code generation, code translation, code comments, and testcase generation, CodeFuse performs better than other models when confronted with Chinese prompts., Comment: Accepted by ICSE-SEIP 2024

Early prediction of Alzheimer's disease (AD) is crucial for timely intervention and treatment. This study aims to use machine learning approaches to analyze longitudinal electronic health records (EHRs) of patients with AD and identify signs and symptoms that can predict AD onset earlier. We used a case-control design with longitudinal EHRs from the U.S. Department of Veterans Affairs Veterans Health Administration (VHA) from 2004 to 2021. Cases were VHA patients with AD diagnosed after 1/1/2016 based on ICD-10-CM codes, matched 1:9 with controls by age, sex and clinical utilization with replacement. We used a panel of AD-related keywords and their occurrences over time in a patient's longitudinal EHRs as predictors for AD prediction with four machine learning models. We performed subgroup analyses by age, sex, and race/ethnicity, and validated the model in a hold-out and "unseen" VHA stations group. Model discrimination, calibration, and other relevant metrics were reported for predictions up to ten years before ICD-based diagnosis. The study population included 16,701 cases and 39,097 matched controls. The average number of AD-related keywords (e.g., "concentration", "speaking") per year increased rapidly for cases as diagnosis approached, from around 10 to over 40, while remaining flat at 10 for controls. The best model achieved high discriminative accuracy (ROCAUC 0.997) for predictions using data from at least ten years before ICD-based diagnoses. The model was well-calibrated (Hosmer-Lemeshow goodness-of-fit p-value = 0.99) and consistent across subgroups of age, sex and race/ethnicity, except for patients younger than 65 (ROCAUC 0.746). Machine learning models using AD-related keywords identified from EHR notes can predict future AD diagnoses, suggesting its potential use for identifying AD risk using EHR notes, offering an affordable way for early screening on large population., Comment: 24 pages

This paper introduces the first attempt to employ a localized meshless method to analyze time-harmonic acoustic wave propagation on curved surfaces with periodic holes/inclusions. In particular, the generalized finite difference method is used as a localized meshless technique to discretize the surface gradient and Laplace-Beltrami operators defined extrinsically in the governing equations. An absorbing boundary condition is introduced to reduce reflections from boundaries and accurately simulate wave propagation on unclosed surfaces with periodic inclusions. Several benchmark examples demonstrate the efficiency and accuracy of the proposed method in simulating acoustic wave propagation on surfaces with diverse geometries, including complex shapes and periodic holes or inclusions., Comment: 24 pages, 17 figures

In this paper, we study the large deviation principle (LDP) for two types (Type I and Type II) of multiplicative Ising models. For Types I and II, the explicit formulas for the free energy functions and the associated rate functions are derived. Furthermore, we prove that those free energy functions are differentiable, which indicates that both systems are characterized by a lack of phase transition phenomena.

An interesting result recently reported for Mrk 421 is the detection of a significant excess at hard X-ray energies, which could provide useful information for investigating particle acceleration and emission mechanisms in the relativistic jet. Considering a two-injection scenario, we develop a self-consistent one-zone leptonic model to understand the origin of the hard X-ray excess in Mrk 421 during the period of extremely low X-ray and very high energy (VHE) flux in 2013 January. In the model, two populations of mono-energetic ultrarelativistic electrons are injected into the emission region that is a magnetized plasmoid propagating along the blazar jet. We numerically calculate the emitting electron energy distribution by solving a kinetic equation that incorporates both shock acceleration and stochastic acceleration processes. Moreover, we infer analytic expressions relating electrons acceleration, cooling, escape and injection to the observed spectra and variability. For the injection luminosity in particular, we derive a new approximate analytical expression for the case of continual injection with a mono-energetic distribution. Based on a comparison between the theoretical predictions and the observed SED, we conclude that the hard X-ray excess observed in Mrk 421 may be due to the synchrotron radiation emitted by an additional electrons population, which is co-spatial with an electron population producing simultaneous Optical/UV, soft X-ray, and $\gamma$-ray emissions. The stochastic acceleration may play a major role in producing the observed X-ray spectrum., Comment: 12 pages, 4 figures, 2 tables, Accepted for publication in ApJ