arXiv subjects

Ze Liu

Publications and source records attributed to Ze Liu.

At least 37 records · Page 2Linked to original sources

arXiv2024

Type-printable photodetector arrays for multichannel meta-infrared imaging

Multichannel meta-imaging, inspired by the parallel-processing capability of neuromorphic computing, offers significant advancements in resolution enhancement and edge discrimination in imaging systems, extending even into the mid- to far-infrared spectrum. Currently typical multichannel infrared imaging systems consist of separating optical gratings or merging multi-cameras, which require complex circuit design and heavy power consumption, hindering the implementation of advanced human-eye-like imagers. Here, we present a novel approach for printable graphene plasmonic photodetector arrays driven by a ferroelectric superdomain for multichannel meta-infrared imaging with enhanced edge discrimination. The fabricated photodetectors exhibited multiple spectral responses with zero-bias operation by directly rescaling the ferroelectric superdomain instead of reconstructing the separated gratings. We also demonstrated enhanced and faster shape classification (98.1%) and edge detection (98.2%) using our multichannel infrared images compared with single-channel detectors. Our proof-of-concept photodetector arrays simplify multichannel infrared imaging systems and hold great potential for applications in efficient edge detection in human-brain-type machine vision.

physics.optics↗

arXiv2024

F5C-finder: An Explainable and Ensemble Biological Language Model for Predicting 5-Formylcytidine Modifications on mRNA

As a prevalent and dynamically regulated epigenetic modification, 5-formylcytidine (f5C) is crucial in various biological processes. However, traditional experimental methods for f5C detection are often laborious and time-consuming, limiting their ability to map f5C sites across the transcriptome comprehensively. While computational approaches offer a cost-effective and high-throughput alternative, no recognition model for f5C has been developed to date. Drawing inspiration from language models in natural language processing, this study presents f5C-finder, an ensemble neural network-based model utilizing multi-head attention for the identification of f5C. Five distinct feature extraction methods were employed to construct five individual artificial neural networks, and these networks were subsequently integrated through ensemble learning to create f5C-finder. 10-fold cross-validation and independent tests demonstrate that f5C-finder achieves state-of-the-art (SOTA) performance with AUC of 0.807 and 0.827, respectively. The result highlights the effectiveness of biological language model in capturing both the order (sequential) and functional meaning (semantics) within genomes. Furthermore, the built-in interpretability allows us to understand what the model is learning, creating a bridge between identifying key sequential elements and a deeper exploration of their biological functions.

q-bio.GN↗

arXiv2024

Nanofabrication beyond optical diffraction limit: Optical driven assembly enabled by superlubricity

The optical manipulation of nanoparticles on superlubricity surfaces is investigated. The research revealed that, due to the near-zero static friction and extremely low dynamic friction at superlubricity interfaces, the maximum intensity for controlling the optical field can be less than 100 W/cm$^2$, which is nine orders of magnitude lower than controlling nanoparticles on traditional interfaces. The controlled nanoparticle radius can be as small as 5 nm, which is more than one order of magnitude smaller than nanoparticles controlled through traditional optical manipulation. Manipulation can be achieved in sub-microsecond to microsecond timescales. Furthermore, the manipulation takes place on solid surfaces and in non-liquid environments, with minimal impact from Brownian motion. By appropriately increasing dynamic friction, controlling light intensity, or reducing pressure, the effects of Brownian motion can be eliminated, allowing for the construction of microstructures with a size as small as 1/75 of the wavelength of light. This enables the control of super-resolution optical microstructures. The optical super-resolution manipulation of nanoparticles on superlubricity surfaces will find important applications in fields such as nanofabrication, photolithography, optical metasurface, and biochemical analysis.

physics.app-ph↗

arXiv2023

FP8-LM: Training FP8 Large Language Models

In this paper, we explore FP8 low-bit data formats for efficient training of large language models (LLMs). Our key insight is that most variables, such as gradients and optimizer states, in LLM training can employ low-precision data formats without compromising model accuracy and requiring no changes to hyper-parameters. Specifically, we propose a new FP8 automatic mixed-precision framework for training LLMs. This framework offers three levels of FP8 utilization to streamline mixed-precision and distributed parallel training for LLMs. It gradually incorporates 8-bit gradients, optimizer states, and distributed learning in an incremental manner. Experiment results show that, during the training of GPT-175B model on H100 GPU platform, our FP8 mixed-precision training framework not only achieved a remarkable 39% reduction in real memory usage but also ran 75% faster than the widely adopted BF16 framework (i.e., Megatron-LM), surpassing the speed of Nvidia Transformer Engine by 37%. This largely reduces the training costs for large foundation models. Furthermore, our FP8 mixed-precision training methodology is generic. It can be seamlessly applied to other tasks such as LLM instruction tuning and reinforcement learning with human feedback, offering savings in fine-tuning expenses. Our FP8 low-precision training framework is open-sourced at {https://github.com/Azure/MS-AMP}{aka.ms/MS.AMP}.

cs.LG↗

arXiv2023

Comparative Analysis of Transfer Learning in Deep Learning Text-to-Speech Models on a Few-Shot, Low-Resource, Customized Dataset

Text-to-Speech (TTS) synthesis using deep learning relies on voice quality. Modern TTS models are advanced, but they need large amount of data. Given the growing computational complexity of these models and the scarcity of large, high-quality datasets, this research focuses on transfer learning, especially on few-shot, low-resource, and customized datasets. In this research, "low-resource" specifically refers to situations where there are limited amounts of training data, such as a small number of audio recordings and corresponding transcriptions for a particular language or dialect. This thesis, is rooted in the pressing need to find TTS models that require less training time, fewer data samples, yet yield high-quality voice output. The research evaluates TTS state-of-the-art model transfer learning capabilities through a thorough technical analysis. It then conducts a hands-on experimental analysis to compare models' performance in a constrained dataset. This study investigates the efficacy of modern TTS systems with transfer learning on specialized datasets and a model that balances training efficiency and synthesis quality. Initial hypotheses suggest that transfer learning could significantly improve TTS models' performance on compact datasets, and an optimal model may exist for such unique conditions. This thesis predicts a rise in transfer learning in TTS as data scarcity increases. In the future, custom TTS applications will favour models optimized for specific datasets over generic, data-intensive ones.

cs.SD↗

arXiv2023

V-DETR: DETR with Vertex Relative Position Encoding for 3D Object Detection

We introduce a highly performant 3D object detector for point clouds using the DETR framework. The prior attempts all end up with suboptimal results because they fail to learn accurate inductive biases from the limited scale of training data. In particular, the queries often attend to points that are far away from the target objects, violating the locality principle in object detection. To address the limitation, we introduce a novel 3D Vertex Relative Position Encoding (3DV-RPE) method which computes position encoding for each point based on its relative position to the 3D boxes predicted by the queries in each decoder layer, thus providing clear information to guide the model to focus on points near the objects, in accordance with the principle of locality. In addition, we systematically improve the pipeline from various aspects such as data normalization based on our understanding of the task. We show exceptional results on the challenging ScanNetV2 benchmark, achieving significant improvements over the previous 3DETR in $\rm{AP}_{25}$/$\rm{AP}_{50}$ from 65.0\%/47.0\% to 77.8\%/66.0\%, respectively. In addition, our method sets a new record on ScanNetV2 and SUN RGB-D datasets.Code will be released at http://github.com/yichaoshen-MS/V-DETR.

cs.CV↗

arXiv2023

Shape-dependent friction scaling laws in twisted layered material interfaces

Static friction induced by moiré superstructure in twisted incommensurate finite layered material interfaces reveals unique double periodicity and lack of scaling with contact size. The underlying mechanism involves compensation of incomplete moiré tiles at the rim of rigid polygonal graphene flakes sliding atop fixed graphene or h-BN substrates. The scaling of friction (or lack thereof) with contact size is found to strongly depend on the shape of the slider and the relative orientation between its edges and the emerging superstructure, partially rationalizing scattered experimental data. With careful consideration of the flake edge orientation, twist angle, and sliding direction along the substrate, one should therefore be able to achieve large-scale superlubricity via shape tailoring.

physics.app-ph↗

arXiv2023

An Adaptive Hybrid Channel Reservation Medium Access Control Protocol for Differentiated QoS

In a densely deployed distributed wireless network, there may be various types of traffic with differentiated Quality of Service (QoS) requirements. However, when the network is heavily loaded, the collision increases significantly, making it difficult to guarantee the QoS of traffic. Designing an efficient Medium Access Control (MAC) protocol to guarantee the QoS of different types of traffic is an essential research direction. Channel reservation mechanism is a promising approach to improving QoS. However, few studies have focused on the channel reservation mechanism for differentiated traffic. It is difficult to take into account both the QoS of real-time traffic and the collision issue for ordinary traffic. To address this issue, this paper proposes the Differentiated Service Guarantee Adaptive Reservation Mechanism (DSGARM) protocol. A hybrid reservation mechanism is proposed by combining the absolute reservation mechanism and the relative reservation mechanism. The absolute reservation mechanism is adopted for real-time traffic. Meanwhile, the relative reservation mechanism is adopted for ordinary traffic. An adaptive algorithm is proposed to calculate the reservation parameters that meet the delay requirements based on the network conditions. The proposed work can be widely applied in the densely deployed distributed wireless network with differentiated QoS requirements. In addition, this paper establishes a mathematical model for the proposed mechanism and theoretically analyzes the performance. Simulations verify that the mathematical model provides a good approximation of the protocol performance and demonstrates the advantages of the proposed protocol.

cs.NI↗

arXiv2023

Tutel: Adaptive Mixture-of-Experts at Scale

Sparsely-gated mixture-of-experts (MoE) has been widely adopted to scale deep learning models to trillion-plus parameters with fixed computational cost. The algorithmic performance of MoE relies on its token routing mechanism that forwards each input token to the right sub-models or experts. While token routing dynamically determines the amount of expert workload at runtime, existing systems suffer inefficient computation due to their static execution, namely static parallelism and pipelining, which does not adapt to the dynamic workload. We present Flex, a highly scalable stack design and implementation for MoE with dynamically adaptive parallelism and pipelining. Flex designs an identical layout for distributing MoE model parameters and input data, which can be leveraged by all possible parallelism or pipelining methods without any mathematical inequivalence or tensor migration overhead. This enables adaptive parallelism/pipelining optimization at zero cost during runtime. Based on this key design, Flex also implements various MoE acceleration techniques. Aggregating all techniques, Flex finally delivers huge speedup at any scale -- 4.96x and 5.75x speedup of a single MoE layer over 16 and 2,048 A100 GPUs, respectively, over the previous state-of-the-art. Our evaluation shows that Flex efficiently and effectively runs a real-world MoE-based model named SwinV2-MoE, built upon Swin Transformer V2, a state-of-the-art computer vision architecture. On efficiency, Flex accelerates SwinV2-MoE, achieving up to 1.55x and 2.11x speedup in training and inference over Fairseq, respectively. On effectiveness, the SwinV2-MoE model achieves superior accuracy in both pre-training and down-stream computer vision tasks such as COCO object detection than the counterpart dense model, indicating the readiness of Flex for end-to-end real-world model training and inference.

cs.DC↗

arXiv2023

An Adaptive Channel Reservation MAC Protocol Based on Forwarding Traffic of Key Nodes

Ad Hoc networks with multi-hop topology are widely used in military and civilian applications. One challenge for Ad Hoc networks is to design efficient Media Access Control (MAC) protocols to ensure the quality of service (QoS). In Ad Hoc networks, there is a kind of node called key node, which undertakes more forwarding traffic than other surrounding nodes. The number of neighbor nodes around key nodes is often large, and the surrounding channel environment and interference are often more complex. Thus, the key nodes can hardly get enough channel access opportunities, resulting in poor end-to-end performance. Therefore, we propose an adaptive channel reservation MAC protocol based on forwarding traffic of key nodes, which is aimed at alleviating the congestion for key nodes. Nodes initiate reservations for future transmission time according to the buffer status before sending packets and then calculate the Weight of Reservation Ability (WRA). The node adaptively adjusts its reservation opportunity by comparing the WRA with neighbor nodes, thus improving the channel access efficiency and ensuring the transmission opportunity of key nodes. Extensive simulation confirms that our proposed FTKN-CRM provides significant improvements in end-to-end performance over the IEEE 802.11ax protocol and other reservation access protocols.

cs.NI↗

arXiv2023

Human Pose as Compositional Tokens

Human pose is typically represented by a coordinate vector of body joints or their heatmap embeddings. While easy for data processing, unrealistic pose estimates are admitted due to the lack of dependency modeling between the body joints. In this paper, we present a structured representation, named Pose as Compositional Tokens (PCT), to explore the joint dependency. It represents a pose by M discrete tokens with each characterizing a sub-structure with several interdependent joints. The compositional design enables it to achieve a small reconstruction error at a low cost. Then we cast pose estimation as a classification task. In particular, we learn a classifier to predict the categories of the M tokens from an image. A pre-learned decoder network is used to recover the pose from the tokens without further post-processing. We show that it achieves better or comparable pose estimation results as the existing methods in general scenarios, yet continues to work well when occlusion occurs, which is ubiquitous in practice. The code and models are publicly available at https://github.com/Gengzigang/PCT.

cs.CV↗

arXiv2022

Registry-Dependent Potential for interfaces of water with graphene

An anisotropic interlayer potential that can accurately describe the van der Waals interaction of the water-graphene interface is presented. The force field is benchmarked against the many-body dispersion-corrected density functional theory. The parameterization of ILP yields good agreement with the reference dataset of binding energy curves and sliding potential energy surfaces for various configurations of a water molecule deposited on monolayer graphene, indicating the developed force field enhancing significantly the accuracy in the empirical description of water-graphene interfacial interactions. The water contact angles of monolayer and multilayer graphene extracted from molecular dynamics simulations based on this force field are close to the experimental measurements and predict the hydrophilic nature of graphene. The theoretical approach proposed in this work can be easily extended to mimic the van der Waals interactions between water molecules and various two-dimensional layered materials for studying their wetting properties.

physics.chem-ph↗

arXiv2022

Could Giant Pretrained Image Models Extract Universal Representations?

Frozen pretrained models have become a viable alternative to the pretraining-then-finetuning paradigm for transfer learning. However, with frozen models there are relatively few parameters available for adapting to downstream tasks, which is problematic in computer vision where tasks vary significantly in input/output format and the type of information that is of value. In this paper, we present a study of frozen pretrained models when applied to diverse and representative computer vision tasks, including object detection, semantic segmentation and video action recognition. From this empirical analysis, our work answers the questions of what pretraining task fits best with this frozen setting, how to make the frozen setting more flexible to various downstream tasks, and the effect of larger model sizes. We additionally examine the upper bound of performance using a giant frozen pretrained model with 3 billion parameters (SwinV2-G) and find that it reaches competitive performance on a varied set of major benchmarks with only one shared frozen base network: 60.0 box mAP and 52.2 mask mAP on COCO object detection test-dev, 57.6 val mIoU on ADE20K semantic segmentation, and 81.7 top-1 accuracy on Kinetics-400 action recognition. With this work, we hope to bring greater attention to this promising path of freezing pretrained image models.

cs.CV↗

arXiv2022

Origin of frictional scaling law in circular twist layered interfaces: simulations and theory

Structural superlubricity based on twisted layered materials has stimulated great research interests. Recent MD simulations show that the circular twisted bilayer graphene (tBLG) presenting a size scaling of friction with strong Moiré-level oscillations. To reveal the physical origin of observed abnormal scaling, we proposed a theoretical formula and derived the analytic expression of frictional size scaling law of tBLG. The predicted twist angle dependent scaling law agrees well with MD simulations and provides a rationalizing explanation for the scattered power scaling law measured in various experiments. Finally, we show clear evidence that the origin of the scaling law comes from the Moiré boundary, that is, the remaining part of the twisted layered interfaces after deleting the internal complete Moiré supercells. Our work provides new physical insights into the friction origin of layered materials and highlights the importance of accounting for Moiré boundary in the thermodynamic models of layered materials.

physics.app-ph↗

arXiv2022

Swin Transformer V2: Scaling Up Capacity and Resolution

Large-scale NLP models have been shown to significantly improve the performance on language tasks with no signs of saturation. They also demonstrate amazing few-shot capabilities like that of human beings. This paper aims to explore large-scale models in computer vision. We tackle three major issues in training and application of large vision models, including training instability, resolution gaps between pre-training and fine-tuning, and hunger on labelled data. Three main techniques are proposed: 1) a residual-post-norm method combined with cosine attention to improve training stability; 2) A log-spaced continuous position bias method to effectively transfer models pre-trained using low-resolution images to downstream tasks with high-resolution inputs; 3) A self-supervised pre-training method, SimMIM, to reduce the needs of vast labeled images. Through these techniques, this paper successfully trained a 3 billion-parameter Swin Transformer V2 model, which is the largest dense vision model to date, and makes it capable of training with images of up to 1,536$\times$1,536 resolution. It set new performance records on 4 representative vision tasks, including ImageNet-V2 image classification, COCO object detection, ADE20K semantic segmentation, and Kinetics-400 video action classification. Also note our training is much more efficient than that in Google's billion-level visual models, which consumes 40 times less labelled data and 40 times less training time. Code is available at \url{https://github.com/microsoft/Swin-Transformer}.

cs.CV↗

arXiv2022

A sampling scheme for estimating the prevalence of a pandemic

The spread of COVID-19 makes it essential to investigate its prevalence. In such investigation research, as far as we know, the widely-used sampling methods didn't use the information sufficiently about the numbers of the previously diagnosed cases, which provides a priori information about the true numbers of infections. This motivates us to develop a new, two-stage sampling method in this paper, which utilises the information about the distributions of both population and diagnosed cases, to investigate the prevalence more efficiently. The global likelihood sampling, a robust and efficient sampler to draw samples from any probability density function, is used in our sampling strategy, and thus, our new method can automatically adapt to the complicated distributions of population and cases. Moreover, the corresponding estimating method is simple, which facilitates the practical implementation. Some recommendations for practical implementation are given. Finally, several simulations and a practical example verified its efficiency.

stat.AP↗

arXiv2021

Thermodynamic model of twisted bilayer graphene: Configuration entropy matters

Twisted bilayer materials have attracted tremendous attention due to their unique and novel properties. Here, we derive a thermodynamic model for twisted bilayer graphene (tBLG) within the framework of the classical statistical mechanics, based on which, the configuration entropy reflecting the number of micro-status in moire unit-cells, is directly derived from the Helmholtz free energy with a clear physical interpretation. More importantly, we show the configuration entropy of tBLG relative to the AB-stacked bilayer graphene is proportional to the logarithmic function of the ratio of moire period and the atomic lattice constant, which we found dominates the Helmholtz free energy of tBLG and can well explain experimental observations in superlubric contacts. Our work provides a theoretical foundation for studying moire effect of incommensurate contact interfaces and could facilitate twisting based applications such as superlubricity.

cond-mat.mes-hall↗

arXiv2021

Swin Transformer: Hierarchical Vision Transformer using Shifted Windows

This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with \textbf{S}hifted \textbf{win}dows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at~\url{https://github.com/microsoft/Swin-Transformer}.

cs.CV↗