arXiv subjects

Tao Li

Publications and source records attributed to Tao Li.

At least 235 records · Page 13Linked to original sources

arXiv2024

Personalized Sleep Staging Leveraging Source-free Unsupervised Domain Adaptation

Sleep staging is crucial for assessing sleep quality and diagnosing related disorders. Recent deep learning models for automatic sleep staging using polysomnography often suffer from poor generalization to new subjects because they are trained and tested on the same labeled datasets, overlooking individual differences. To tackle this issue, we propose a novel Source-Free Unsupervised Individual Domain Adaptation (SF-UIDA) framework. This two-step adaptation scheme allows the model to effectively adjust to new unlabeled individuals without needing source data, facilitating personalized customization in clinical settings. Our framework has been applied to three established sleep staging models and tested on three public datasets, achieving state-of-the-art performance.

cs.LG↗

arXiv2024

PandaX-xT: a Multi-ten-tonne Liquid Xenon Observatory at the China Jinping Underground Laboratory

We propose a major upgrade to the existing PandaX-4T experiment in the China Jinping Underground Laboratory. The new experiment, PandaX-xT, will be a multi-ten-tonne liquid xenon, ultra-low background, and general-purpose observatory. The full-scaled PandaX-xT contains a 43-tonne liquid xenon active target. Such an experiment will significantly advance our fundamental understanding of particle physics and astrophysics. The sensitivity of dark matter direct detection will be improved by nearly two orders of magnitude compared to the current best limits, approaching the so-called "neutrino floor" for a dark matter mass above 10 GeV/$c^2$, providing a decisive test to the Weakly Interacting Massive Particle paradigm. By searching for the neutrinoless double beta decay of $^{136}$Xe isotope in the detector, the effective Majorana neutrino mass can be measured to a [10 -- 41] meV/$c^2$ sensitivity, providing a key test to the Dirac/Majorana nature of neutrino s. Astrophysical neutrinos and other ultra-rare interactions can also be measured and searched for with an unprecedented background level, opening up new windows of discovery. Depending on the findings, PandaX-xT will seek the next stage upgrade utilizing isotopic separation on natural xenon.

hep-ex↗

arXiv2024

Machine learning enhanced multi-particle tracking in solid fuel combustion

Particle velocimetry is essential in solid fuel combustion studies, however, the accurate detection and tracking of particles in high Particle Number Density (PND) combustion scenario remain challenging. The current study advances the machine-learning approaches for precise velocity measurements of solid particles. For this, laser imaging experiments were performed for high-volatile bituminous coal particles burning in a laminar flow reactor. Particle positions were imaged using time-resolved Mie scattering. Various detection methods, including conventional blob detection and Machine Learning (ML) based You Only Look Once (YOLO) and Realtime Detection Transformer (RT-DETR) were employed and bench marked.~Particle tracking was performed using the Simple Online Realtime Tracking (SORT) algorithm. The results demonstrated the capability of machine learning models trained on low-PND data for prediction of high-PND data. Slicing Aided Hyper Inference (SAHI) algorithm is important for the better performance of the used models. By evaluating the velocity statistics, it is found that the mean particle velocity decreases with increasing PND, primarily due to stronger particle interactions. The particle dynamics are closely related to the position of combustion zone observed in the previous study. Thus, PND is considered as the dominant factor for the particle group combustion behavior of high-volatile solid fuels.

physics.app-ph↗

arXiv2024

Heralded nonlocal quantum gates for distributed quantum computation in a decoherence-free subspace

We propose a heralded protocol for implementing nontrivial quantum gates on two stationary qubits coupled to spatially separated cavities. By dynamically controlling the evolution of the composite system, nonlocal two-qubit quantum (e.g., CPHASE and CNOT) gates can be achieved without real excitations of either cavity modes or atoms. The success of our protocol is conditioned on projecting an auxiliary atom onto a postselected state, which simultaneously removes various detrimental effects of dissipation on the gate fidelity. In principle, the success probability of the gate can approach unity as the single-atom cooperativity becomes sufficiently large.Furthermore, we show its application for implementing single- and two-qubit gates within a decoherence-free subspace that is immune to a collective dephasing noise. This faithful, heralded, and nonlocal protocol could, therefore, be useful for distributed quantum computation and scalable quantum networks.

quant-ph↗

arXiv2024

Spatio-temporal cooperative control Method of Highway Ramp Merge Based on Vehicle-road Coordination

The merging area of highway ramps faces multiple challenges, including traffic congestion, collision risks, speed mismatches, driver behavior uncertainties, limited visibility, and bottleneck effects. However, autonomous vehicles engaging in depth coordination between vehicle and road in merging zones, by pre-planning and uploading travel trajectories, can significantly enhance the safety and efficiency of merging zones.In this paper,we mainly introduce mainline priority cooperation method to achieve the time and space cooperative control of highway merge.Vehicle-mounted intelligent units share real-time vehicle status and driving intentions with Road Section Management Units, which pre-plan the spatiotemporal trajectories of vehicle travel. After receiving these trajectories, Vehicle Intelligent Units strictly adhere to them. Through this deep collaboration between vehicles and roads, conflicts in time and space during vehicle travel are eliminated in advance.

eess.SY↗

arXiv2024

Digital Twin-Enhanced Wireless Indoor Navigation: Achieving Efficient Environment Sensing with Zero-Shot Reinforcement Learning

Millimeter-wave (mmWave) communication is a vital component of future generations of mobile networks, offering not only high data rates but also precise beams, making it ideal for indoor navigation in complex environments. However, the challenges of multipath propagation and noisy signal measurements in indoor spaces complicate the use of mmWave signals for navigation tasks. Traditional physics-based methods, such as following the angle of arrival (AoA), often fall short in complex scenarios, highlighting the need for more sophisticated approaches. Digital twins, as virtual replicas of physical environments, offer a powerful tool for simulating and optimizing mmWave signal propagation in such settings. By creating detailed, physics-based models of real-world spaces, digital twins enable the training of machine learning algorithms in virtual environments, reducing the costs and limitations of physical testing. Despite their advantages, current machine learning models trained in digital twins often overfit specific virtual environments and require costly retraining when applied to new scenarios. In this paper, we propose a Physics-Informed Reinforcement Learning (PIRL) approach that leverages the physical insights provided by digital twins to shape the reinforcement learning (RL) reward function. By integrating physics-based metrics such as signal strength, AoA, and path reflections into the learning process, PIRL enables efficient learning and improved generalization to new environments without retraining. Our experiments demonstrate that the proposed PIRL, supported by digital twin simulations, outperforms traditional heuristics and standard RL models, achieving zero-shot generalization in unseen environments and offering a cost-effective, scalable solution for wireless indoor navigation.

cs.RO↗

arXiv2024

Inference-to-complete: A High-performance and Programmable Data-plane Co-processor for Neural-network-driven Traffic Analysis

Neural-networks-driven intelligent data-plane (NN-driven IDP) is becoming an emerging topic for excellent accuracy and high performance. Meanwhile we argue that NN-driven IDP should satisfy three design goals: the flexibility to support various NNs models, the low-latency-high-throughput inference performance, and the data-plane-unawareness harming no performance and functionality. Unfortunately, existing work either over-modify NNs for IDP, or insert inline pipelined accelerators into the data-plane, failing to meet the flexibility and unawareness goals. In this paper, we propose Kaleidoscope, a flexible and high-performance co-processor located at the bypass of the data-plane. To address the challenge of meeting three design goals, three key techniques are presented. The programmable run-to-completion accelerators are developed for flexible inference. To further improve performance, we design a scalable inference engine which completes low-latency and low-cost inference for the mouse flows, and perform complex NNs with high-accuracy for the elephant flows. Finally, raw-bytes-based NNs are introduced, which help to achieve unawareness. We prototype Kaleidoscope on both FPGA and ASIC library. In evaluation on six NNs models, Kaleidoscope reaches 256-352 ns inference latency and 100 Gbps throughput with negligible influence on the data-plane. The on-board tested NNs perform state-of-the-art accuracy among other NN-driven IDP, exhibiting the the significant impact of flexibility on enhancing traffic analysis accuracy.

cs.NI↗

arXiv2024

Tunnel number one knots satisfy the Berge Conjecture

Let $K$ be a tunnel number one knot in $M$ with irreducible knot exterior, where $M$ is either $S^3$, or a connected sum of $S^2\times S^1$ with any lens space. (In particular, this includes $M = S^2\times S^1$.) We prove that if a non-trivial Dehn surgery on $K$ yields a lens space, then $K$ is a doubly primitive knot in $M$. For $M = S^3$ this resolves the tunnel number one Berge Conjecture. For $M = S^2\times S^1$ this resolves a conjecture of Greene and Baker-Buck-Lecuona for tunnel number one knots.

math.GT↗

arXiv2024

Learning from Response not Preference: A Stackelberg Approach for LLM Detoxification using Non-parallel Data

Text detoxification, a variant of style transfer tasks, finds useful applications in online social media. This work presents a fine-tuning method that only uses non-parallel data to turn large language models (LLM) into a detoxification rewritter. We model the fine-tuning process as a Stackelberg game between an LLM (leader) and a toxicity screener (follower), which is a binary style classifier (toxic or non-toxic). The LLM aims to align its preference according to the screener and generate paraphases passing the screening. The primary challenge of non-parallel data fine-tuning is incomplete preference. In the case of unsuccessful paraphrases, the classifier cannot establish a preference between the input and paraphrase, as they belong to the same toxic style. Hence, preference-alignment fine-tuning methods, such as direct preference optimization (DPO), no longer apply. To address the challenge of incomplete preference, we propose Stackelberg response optimization (SRO), adapted from DPO, to enable the LLM to learn from the follower's response. The gist is that SRO decreases the likelihood of generating the paraphrase if it fails the follower's screening while performing DPO on the pair of the toxic input and its paraphrase when the latter passes the screening. Experiments indicate that the SRO-fine-tunned LLM achieves satisfying performance comparable to state-of-the-art models regarding style accuracy, content similarity, and fluency. The overall detoxification performance surpasses other computing methods and matches the human reference. Additional empirical evidence suggests that SRO is sensitive to the screener's feedback, and a slight perturbation leads to a significant performance drop. We release the code and LLM models at \url{https://github.com/XXXinhong/Detoxification_LLM}.

cs.CL↗

arXiv2024

Resfusion: Denoising Diffusion Probabilistic Models for Image Restoration Based on Prior Residual Noise

Recently, research on denoising diffusion models has expanded its application to the field of image restoration. Traditional diffusion-based image restoration methods utilize degraded images as conditional input to effectively guide the reverse generation process, without modifying the original denoising diffusion process. However, since the degraded images already include low-frequency information, starting from Gaussian white noise will result in increased sampling steps. We propose Resfusion, a general framework that incorporates the residual term into the diffusion forward process, starting the reverse process directly from the noisy degraded images. The form of our inference process is consistent with the DDPM. We introduced a weighted residual noise, named resnoise, as the prediction target and explicitly provide the quantitative relationship between the residual term and the noise term in resnoise. By leveraging a smooth equivalence transformation, Resfusion determine the optimal acceleration step and maintains the integrity of existing noise schedules, unifying the training and inference processes. The experimental results demonstrate that Resfusion exhibits competitive performance on ISTD dataset, LOL dataset and Raindrop dataset with only five sampling steps. Furthermore, Resfusion can be easily applied to image generation and emerges with strong versatility. Our code and model are available at https://github.com/nkicsl/Resfusion.

cs.CV↗

arXiv2024

Synthesis and stability of biomolecules in C-H-O-N fluids under Earth's upper mantle conditions

How life started on Earth is an unsolved mystery. There are various hypotheses for the location ranging from outer space to the seafloor, subseafloor or potentially deeper. Here, we applied extensive ab initio molecular dynamics (AIMD) simulations to study chemical reactions between NH$_3$, H$_2$O, H$_2$, and CO at pressures (P) and temperatures (T) approximating the conditions of Earth's upper mantle (i.e. 10-13 GPa, 1000-1400 K). Contrary to the previous assumptions that larger organic molecules might readily disintegrate in aqueous solutions at extreme P-T conditions, we found that many organic compounds formed without any catalysts and persisted in C-H-O-N fluids under these extreme conditions, including glycine, ribose, urea, and uracil-like molecules. Particularly, our free energy calculations showed that the C-N bond is thermodynamically stable at 10 GPa and 1400 K. Moreover, while the pyranose (six-membered-ring) form of ribose is more stable than the furanose (five-membered-ring) form at ambient conditions, we observed the predominant formation of the five-membered-ring form of ribose at extreme conditions, which is consistent with the exclusive incorporation of $β$-D-ribofuranose in RNA. We have uncovered a previously unexplored pathway through which the crucial biomolecules could be abiotically synthesized from geofluids in the deep interior of Earth and other planets and these formed biomolecules could potentially contribute to the early stage of the emergency of life.

physics.bio-ph↗

arXiv2024

Influential Language Data Selection via Gradient Trajectory Pursuit

Curating a desirable dataset for training has been the core of building highly capable large language models (Touvron et al., 2023; Achiam et al., 2023; Team et al.,2024). Gradient influence scores (Pruthi et al., 2020; Xia et al., 2024) are shown to be correlated with model performance and are commonly used as the criterion for data selection. However, existing methods are built upon either individual sample rankings or inefficient matching process, leading to suboptimal performance or scaling up issues.In this paper, we propose Gradient Trajectory Pursuit (GTP), an algorithm that performs pursuit of gradient trajectories via jointly selecting data points under an L0-norm regularized objective. The proposed algorithm highlights: (1) joint selection instead of independent top-k selection, which automatically de-duplicates samples; (2) higher efficiency with compressive sampling processes, which can be further sped up using a distributed framework. In the experiments, we demonstrate the algorithm in both in-domain and target-domain selection benchmarks and show that it outperforms top-k selection and competitive algorithms consistently, for example, our algorithm chooses as low as 0.5% data to achieve full performance on the targeted instruction tuning tasks

cs.AI↗

arXiv2024

Meta Stackelberg Game: Robust Federated Learning against Adaptive and Mixed Poisoning Attacks

Federated learning (FL) is susceptible to a range of security threats. Although various defense mechanisms have been proposed, they are typically non-adaptive and tailored to specific types of attacks, leaving them insufficient in the face of multiple uncertain, unknown, and adaptive attacks employing diverse strategies. This work formulates adversarial federated learning under a mixture of various attacks as a Bayesian Stackelberg Markov game, based on which we propose the meta-Stackelberg defense composed of pre-training and online adaptation. {The gist is to simulate strong attack behavior using reinforcement learning (RL-based attacks) in pre-training and then design meta-RL-based defense to combat diverse and adaptive attacks.} We develop an efficient meta-learning approach to solve the game, leading to a robust and adaptive FL defense. Theoretically, our meta-learning algorithm, meta-Stackelberg learning, provably converges to the first-order $\varepsilon$-meta-equilibrium point in $O(\varepsilon^{-2})$ gradient iterations with $O(\varepsilon^{-4})$ samples per iteration. Experiments show that our meta-Stackelberg framework performs superbly against strong model poisoning and backdoor attacks of uncertain and unknown types.

cs.LG↗

arXiv2024

Reactor neutrino liquid xenon coherent elastic scattering experiment

Coherent elastic neutrino-nucleus scattering (CEvNS) provides a unique probe for neutrino properties Beyond the Standard Model (BSM) physics. REactor neutrino LIquid xenon Coherent Scattering experiment (RELICS), a proposed reactor neutrino program using liquid xenon time projection chamber (LXeTPC) technology, aims to investigate the CEvNS process of antineutrinos off xenon atomic nuclei. In this work, the design of the experiment is studied and optimized based on Monte Carlo (MC) simulations. To achieve a sufficiently low energy threshold for CEvNS detection, an ionization-only analysis channel is adopted for RELICS. A high emission rate of delayed electrons after a big ionization signal is the major background, leading to an analysis threshold of 120 photo-electrons in the CEvNS search. The second largest background, nuclear recoils induced by cosmic-ray neutrons, is suppressed via a passive water shield. The physics potential of RELICS is explored with a 32 kg*yr exposure at a baseline of 25 m from a reactor core with a 3 GW thermal power. In an energy range of 120 to 300 PE, corresponding to an average nuclear recoil from 0.63 to 1.36 keV considering the liquid xenon response and detector-related effect, we expect 4639.7 CEvNS and 1687.8 background events. The sensitivity of RELICS to the weak mixing angle is investigated at a low momentum transfer. Our study shows that RELICS can further improve the constraints on the non-standard neutrino interaction (NSI) compared to the current best results.

hep-ex↗

arXiv2024

PromptIntern: Saving Inference Costs by Internalizing Recurrent Prompt during Large Language Model Fine-tuning

Recent advances in fine-tuning large language models (LLMs) have greatly enhanced their usage in domain-specific tasks. Despite the success, fine-tuning continues to rely on repeated and lengthy prompts, which escalate computational expenses, require more resources, and lead to slower inference. In this paper, we present a novel approach, PromptIntern, which internalizes prompt knowledge during model fine-tuning to achieve efficient inference and save costs. Instead of compressing the prompts for a vanilla model, PromptIntern aims to embed the recurrent prompt directly into the model parameters. We design a fine-tuning pipeline that includes instruction template compression, few-shot example absorption, and a progressive internalization strategy, effectively diminishing the need for intricate prompts during inference. Comprehensive experiments on challenging NL2Code tasks demonstrate that our method reduces input tokens by more than 90%, accelerates inference by 4.2 times, and reduces monetary inference costs by 88.3%.

cs.CL↗

arXiv2024

ADFQ-ViT: Activation-Distribution-Friendly Post-Training Quantization for Vision Transformers

Vision Transformers (ViTs) have exhibited exceptional performance across diverse computer vision tasks, while their substantial parameter size incurs significantly increased memory and computational demands, impeding effective inference on resource-constrained devices. Quantization has emerged as a promising solution to mitigate these challenges, yet existing methods still suffer from significant accuracy loss at low-bit. We attribute this issue to the distinctive distributions of post-LayerNorm and post-GELU activations within ViTs, rendering conventional hardware-friendly quantizers ineffective, particularly in low-bit scenarios. To address this issue, we propose a novel framework called Activation-Distribution-Friendly post-training Quantization for Vision Transformers, ADFQ-ViT. Concretely, we introduce the Per-Patch Outlier-aware Quantizer to tackle irregular outliers in post-LayerNorm activations. This quantizer refines the granularity of the uniform quantizer to a per-patch level while retaining a minimal subset of values exceeding a threshold at full-precision. To handle the non-uniform distributions of post-GELU activations between positive and negative regions, we design the Shift-Log2 Quantizer, which shifts all elements to the positive region and then applies log2 quantization. Moreover, we present the Attention-score enhanced Module-wise Optimization which adjusts the parameters of each quantizer by reconstructing errors to further mitigate quantization error. Extensive experiments demonstrate ADFQ-ViT provides significant improvements over various baselines in image classification, object detection, and instance segmentation tasks at 4-bit. Specifically, when quantizing the ViT-B model to 4-bit, we achieve a 10.23% improvement in Top-1 accuracy on the ImageNet dataset.

cs.CV↗

arXiv2024

Beyond Perplexity: Multi-dimensional Safety Evaluation of LLM Compression

Increasingly, model compression techniques enable large language models (LLMs) to be deployed in real-world applications. As a result of this momentum towards local deployment, compressed LLMs will interact with a large population. Prior work on compression typically prioritize preserving perplexity, which is directly analogous to training loss. The impact of compression method on other critical aspects of model behavior\, -- \,particularly safety\, -- \,requires systematic assessment. To this end, we investigate the impact of model compression along four dimensions: (1) degeneration harm, i.e., bias and toxicity in generation; (2) representational harm, i.e., biases in discriminative tasks; (3) dialect bias; and(4) language modeling and downstream task performance. We examine a wide spectrum of LLM compression techniques, including unstructured pruning, semi-structured pruning, and quantization. Our analysis reveals that compression can lead to unexpected consequences. Although compression may unintentionally alleviate LLMs' degeneration harm, it can still exacerbate representational harm. Furthermore, increasing compression produces a divergent impact on different protected groups. Finally, different compression methods have drastically different safety impacts: for example, quantization mostly preserves bias while pruning degrades quickly. Our findings underscore the importance of integrating safety assessments into the development of compressed LLMs to ensure their reliability across real-world applications.\footnote{Our implementation and results are available here: \url{https://github.com/zhichaoxu-shufe/Beyond-Perplexity-Compression-Safety-Eval}}

cs.CL↗

arXiv2024

DeltaDQ: Ultra-High Delta Compression for Fine-Tuned LLMs via Group-wise Dropout and Separate Quantization

Large language models achieve exceptional performance on various downstream tasks through supervised fine-tuning. However, the diversity of downstream tasks and practical requirements makes deploying multiple full-parameter fine-tuned models challenging. Current methods that compress the delta weight struggle to achieve ultra-high compression, failing to minimize the deployment overhead. To address the above issue, we propose a novel distribution-driven delta compression framework DeltaDQ, which utilizes Group-wise Dropout and Separate Quantization to achieve ultra-high compression for the delta weight. We have observed that the matrix-computed intermediate results for the delta weight exhibit extremely small variance and min-max range characteristics, referred to as Balanced Intermediate Results. Exploiting this phenomenon, we introduce Group-wise Dropout to perform dropout on the delta weight using an optimal group size. Furthermore, using Separate Quantization, sparse weights are quantized and decomposed to achieve a lower bit. Experimental results show that DeltaDQ achieves 16x compression with improved accuracy compared to baselines for WizardMath and WizardCoder models across different parameter scales. Moreover, DeltaDQ demonstrates the ability for ultra-high compression ratio, achieving 128x compression for the WizardMath-7B model and 512x compression for the WizardMath-70B model.

cs.LG↗