arXiv subjects

Zeyu Liu

Publications and source records attributed to Zeyu Liu.

At least 109 records · Page 6Linked to original sources

arXiv2023

Signature of Scramblon Effective Field Theory in Random Spin Models

Information scrambling refers to the propagation of information throughout a quantum system. Its study not only contributes to our understanding of thermalization but also has wide implications in quantum information and black hole physics. Recent studies suggest that information scrambling is mediated by collective modes called scramblons. However, a criterion for the validity of scramblon theory in a specific model is still missing. In this work, we address this issue by investigating the signature of the scramblon effective theory in random spin models with all-to-all interactions. We demonstrate that, in scenarios where the scramblon description holds, the late-time operator size distribution can be predicted from its early-time value, requiring no free parameters. As an illustration, we examine whether Brownian circuits exhibit a scramblon description and obtain a positive confirmation both analytically and numerically. We also discuss the prediction of multiple-quantum coherence when the scramblon description is valid. Our findings provide a concrete experimental framework for unraveling the scramblon field theory in random spin models using quantum simulators.

quant-ph↗

arXiv2023

FeatureBooster: Boosting Feature Descriptors with a Lightweight Neural Network

We introduce a lightweight network to improve descriptors of keypoints within the same image. The network takes the original descriptors and the geometric properties of keypoints as the input, and uses an MLP-based self-boosting stage and a Transformer-based cross-boosting stage to enhance the descriptors. The boosted descriptors can be either real-valued or binary ones. We use the proposed network to boost both hand-crafted (ORB, SIFT) and the state-of-the-art learning-based descriptors (SuperPoint, ALIKE) and evaluate them on image matching, visual localization, and structure-from-motion tasks. The results show that our method significantly improves the performance of each task, particularly in challenging cases such as large illumination changes or repetitive patterns. Our method requires only 3.2ms on desktop GPU and 27ms on embedded GPU to process 2000 features, which is fast enough to be applied to a practical system. The code and trained weights are publicly available at github.com/SJTU-ViSYS/FeatureBooster.

cs.CV↗

arXiv2022

In-Sensor & Neuromorphic Computing are all you need for Energy Efficient Computer Vision

Due to the high activation sparsity and use of accumulates (AC) instead of expensive multiply-and-accumulates (MAC), neuromorphic spiking neural networks (SNNs) have emerged as a promising low-power alternative to traditional DNNs for several computer vision (CV) applications. However, most existing SNNs require multiple time steps for acceptable inference accuracy, hindering real-time deployment and increasing spiking activity and, consequently, energy consumption. Recent works proposed direct encoding that directly feeds the analog pixel values in the first layer of the SNN in order to significantly reduce the number of time steps. Although the overhead for the first layer MACs with direct encoding is negligible for deep SNNs and the CV processing is efficient using SNNs, the data transfer between the image sensors and the downstream processing costs significant bandwidth and may dominate the total energy. To mitigate this concern, we propose an in-sensor computing hardware-software co-design framework for SNNs targeting image recognition tasks. Our approach reduces the bandwidth between sensing and processing by 12-96x and the resulting total energy by 2.32x compared to traditional CV processing, with a 3.8% reduction in accuracy on ImageNet.

cs.CV↗

arXiv2022

Hoyer regularizer is all you need for ultra low-latency spiking neural networks

Spiking Neural networks (SNN) have emerged as an attractive spatio-temporal computing paradigm for a wide range of low-power vision tasks. However, state-of-the-art (SOTA) SNN models either incur multiple time steps which hinder their deployment in real-time use cases or increase the training complexity significantly. To mitigate this concern, we present a training framework (from scratch) for one-time-step SNNs that uses a novel variant of the recently proposed Hoyer regularizer. We estimate the threshold of each SNN layer as the Hoyer extremum of a clipped version of its activation map, where the clipping threshold is trained using gradient descent with our Hoyer regularizer. This approach not only downscales the value of the trainable threshold, thereby emitting a large number of spikes for weight update with a limited number of iterations (due to only one time step) but also shifts the membrane potential values away from the threshold, thereby mitigating the effect of noise that can degrade the SNN accuracy. Our approach outperforms existing spiking, binary, and adder neural networks in terms of the accuracy-FLOPs trade-off for complex image recognition tasks. Downstream experiments on object detection also demonstrate the efficacy of our approach.

cs.CV↗

arXiv2022

Artificial Intelligence Security Competition (AISC)

The security of artificial intelligence (AI) is an important research area towards safe, reliable, and trustworthy AI systems. To accelerate the research on AI security, the Artificial Intelligence Security Competition (AISC) was organized by the Zhongguancun Laboratory, China Industrial Control Systems Cyber Emergency Response Team, Institute for Artificial Intelligence, Tsinghua University, and RealAI as part of the Zhongguancun International Frontier Technology Innovation Competition (https://www.zgc-aisc.com/en). The competition consists of three tracks, including Deepfake Security Competition, Autonomous Driving Security Competition, and Face Recognition Security Competition. This report will introduce the competition rules of these three tracks and the solutions of top-ranking teams in each track.

cs.CR↗

arXiv2022

Enabling ISP-less Low-Power Computer Vision

In order to deploy current computer vision (CV) models on resource-constrained low-power devices, recent works have proposed in-sensor and in-pixel computing approaches that try to partly/fully bypass the image signal processor (ISP) and yield significant bandwidth reduction between the image sensor and the CV processing unit by downsampling the activation maps in the initial convolutional neural network (CNN) layers. However, direct inference on the raw images degrades the test accuracy due to the difference in covariance of the raw images captured by the image sensors compared to the ISP-processed images used for training. Moreover, it is difficult to train deep CV models on raw images, because most (if not all) large-scale open-source datasets consist of RGB images. To mitigate this concern, we propose to invert the ISP pipeline, which can convert the RGB images of any dataset to its raw counterparts, and enable model training on raw images. We release the raw version of the COCO dataset, a large-scale benchmark for generic high-level vision tasks. For ISP-less CV systems, training on these raw images result in a 7.1% increase in test accuracy on the visual wake works (VWW) dataset compared to relying on training with traditional ISP-processed RGB datasets. To further improve the accuracy of ISP-less CV models and to increase the energy and bandwidth benefits obtained by in-sensor/in-pixel computing, we propose an energy-efficient form of analog in-pixel demosaicing that may be coupled with in-pixel CNN computations. When evaluated on raw images captured by real sensors from the PASCALRAW dataset, our approach results in a 8.1% increase in mAP. Lastly, we demonstrate a further 20.5% increase in mAP by using a novel application of few-shot learning with thirty shots each for the novel PASCALRAW dataset, constituting 3 classes.

cs.CV↗

arXiv2022

Leveraging Low-Fidelity Data to Improve Machine Learning of Sparse High-Fidelity Thermal Conductivity Data via Transfer Learning

Lattice thermal conductivity (TC) of semiconductors is crucial for various applications, ranging from microelectronics to thermoelectrics. Data-driven approach can potentially establish the critical composition-property relationship needed for fast screening of candidates with desirable TC, but the small number of available data remains the main challenge. TC can be efficiently calculated using empirical models, but they have inferior accuracy compared to the more resource-demanding first-principles calculations. Here, we demonstrate the use of transfer learning (TL) to improve the machine learning models trained on small but high-fidelity TC data from experiments and first-principles calculations, by leveraging a large but low-fidelity data generated from empirical TC models, where the trainings on high- and low-fidelity TC data are treated as different but related tasks. TL improves the model accuracy by as much as 23% in R2 and reduces the average factor difference by as much as 30%. Using the TL model, a large semiconductor database is screened, and several candidates with room temperature TC > 350 W/mK are identified and further verified using first-principles simulations. This study demonstrates that TL can leverage big low-fidelity data as a proxy task to improve models for the target task with high-fidelity but small data. Such a capability of TL may have important implications to materials informatics in general.

cond-mat.mtrl-sci↗

arXiv2022

Ballistic Thermal Transport at Sub-10 nm Laser-Induced Hot Spots in GaN Crystal

Gallium nitride (GaN) is a typical wide-bandgap semiconductor with a critical role in a wide range of electronic applications. Ballistic thermal transport at nanoscale hotspots will greatly reduce the performance of a device when its characteristic length reaches the nanometer scale, due to heat dissipation. In this work, we developed a tip-enhanced Raman thermometry approach to study ballistic thermal transport within the range of 10 nm in GaN, simultaneously achieving laser heating and measuring the local temperature. The Raman results showed that the temperature increase from an Au-coated tip-focused hotspot was up to two times higher (40 K) than that in a bare tip-focused region (20 K). To further investigate the possible mechanisms behind this temperature difference, we performed electromagnetic simulations to generate a highly focused heating field, and observed a highly localized optical penetration, within a range of 10 nm. The phonon mean free path (MFP) of the GaN substrate could thus be determined by comparing the numerical simulation results with the experimentally measured temperature increase which was in good agreement with the average MFP weighted by the mode-specific thermal conductivity, as calculated from first-principles simulations. Our results demonstrate that the phonon MFP of a material can be rapidly predicted through a combination of experiments and simulations, which can find wide application in the thermal management of GaN-based electronics.

cond-mat.mes-hall↗

arXiv2022

Synthesizing Diverse and Physically Stable Grasps with Arbitrary Hand Structures using Differentiable Force Closure Estimator

Existing grasp synthesis methods are either analytical or data-driven. The former one is oftentimes limited to specific application scope. The latter one depends heavily on demonstrations, thus suffers from generalization issues; e.g., models trained with human grasp data would be difficult to transfer to 3-finger grippers. To tackle these deficiencies, we formulate a fast and differentiable force closure estimation method, capable of producing diverse and physically stable grasps with arbitrary hand structures, without any training data. Although force closure has commonly served as a measure of grasp quality, it has not been widely adopted as an optimization objective for grasp synthesis primarily due to its high computational complexity; in comparison, the proposed differentiable method can test a force closure within milliseconds. In experiments, we validate the proposed method's efficacy in 6 different settings.

cs.RO↗

arXiv2022

De Rham prismatic crystals over $\mathcal{O}_K$

We study de Rham prismatic crystals on $(\mathcal{O}_K)_{\bboldΔ}$. We show that a de Rham crystal is controlled by a sequence of matrices $\{A_{m,1}\}_{m \geq 0}$ with $A_{0,1}$ "nilpotent". Using this, we prove that the natural functor from de Rham crystals over $(\mathcal{O}_K)_{\bboldΔ}$ to the category of nearly de Rham representations is fully faithful. The key ingredient is a Sen style decompletion theorem for de Rham representations of $G_K$.

math.NT↗

arXiv2022

P2M-DeTrack: Processing-in-Pixel-in-Memory for Energy-efficient and Real-Time Multi-Object Detection and Tracking

Today's high resolution, high frame rate cameras in autonomous vehicles generate a large volume of data that needs to be transferred and processed by a downstream processor or machine learning (ML) accelerator to enable intelligent computing tasks, such as multi-object detection and tracking. The massive amount of data transfer incurs significant energy, latency, and bandwidth bottlenecks, which hinders real-time processing. To mitigate this problem, we propose an algorithm-hardware co-design framework called Processing-in-Pixel-in-Memory-based object Detection and Tracking (P2M-DeTrack). P2M-DeTrack is based on a custom faster R-CNN-based model that is distributed partly inside the pixel array (front-end) and partly in a separate FPGA/ASIC (back-end). The proposed front-end in-pixel processing down-samples the input feature maps significantly with judiciously optimized strided convolution and pooling. Compared to a conventional baseline design that transfers frames of RGB pixels to the back-end, the resulting P2M-DeTrack designs reduce the data bandwidth between sensor and back-end by up to 24x. The designs also reduce the sensor and total energy (obtained from in-house circuit simulations at Globalfoundries 22nm technology node) per frame by 5.7x and 1.14x, respectively. Lastly, they reduce the sensing and total frame latency by an estimated 1.7x and 3x, respectively. We evaluate our approach on the multi-object object detection (tracking) task of the large-scale BDD100K dataset and observe only a 0.5% reduction in the mean average precision (0.8% reduction in the identification F1 score) compared to the state-of-the-art.

eess.IV↗

arXiv2022

Risk Assessment with Generic Energy Storage under Exogenous and Endogenous Uncertainty

Current risk assessment ignores the stochastic nature of energy storage availability itself and thus lead to potential risk during operation. This paper proposes the redefinition of generic energy storage (GES) that is allowed to offer probabilistic reserve. A data-driven unified model with exogenous and endogenous uncertainty (EXU & EDU) description is presented for four typical types of GES. Moreover, risk indices are proposed to assess the impact of overlooking (EXU & EDU) of GES. Comparative results between EXU & EDU are illustrated in distribution system with day-ahead chance-constrained optimization (CCO) and more severe risks are observed for the latter, which indicate that system operator (SO) should adopt novel strategies for EDU uncertainty.

q-fin.RM↗

arXiv2021

StructDepth: Leveraging the structural regularities for self-supervised indoor depth estimation

Self-supervised monocular depth estimation has achieved impressive performance on outdoor datasets. Its performance however degrades notably in indoor environments because of the lack of textures. Without rich textures, the photometric consistency is too weak to train a good depth network. Inspired by the early works on indoor modeling, we leverage the structural regularities exhibited in indoor scenes, to train a better depth network. Specifically, we adopt two extra supervisory signals for self-supervised training: 1) the Manhattan normal constraint and 2) the co-planar constraint. The Manhattan normal constraint enforces the major surfaces (the floor, ceiling, and walls) to be aligned with dominant directions. The co-planar constraint states that the 3D points be well fitted by a plane if they are located within the same planar region. To generate the supervisory signals, we adopt two components to classify the major surface normal into dominant directions and detect the planar regions on the fly during training. As the predicted depth becomes more accurate after more training epochs, the supervisory signals also improve and in turn feedback to obtain a better depth model. Through extensive experiments on indoor benchmark datasets, the results show that our network outperforms the state-of-the-art methods. The source code is available at https://github.com/SJTU-ViSYS/StructDepth .

cs.CV↗

arXiv2020

Every Corporation Owns Its Image: Corporate Credit Ratings via Convolutional Neural Networks

Credit rating is an analysis of the credit risks associated with a corporation, which reflect the level of the riskiness and reliability in investing. There have emerged many studies that implement machine learning techniques to deal with corporate credit rating. However, the ability of these models is limited by enormous amounts of data from financial statement reports. In this work, we analyze the performance of traditional machine learning models in predicting corporate credit rating. For utilizing the powerful convolutional neural networks and enormous financial data, we propose a novel end-to-end method, Corporate Credit Ratings via Convolutional Neural Networks, CCR-CNN for brevity. In the proposed model, each corporation is transformed into an image. Based on this image, CNN can capture complex feature interactions of data, which are difficult to be revealed by previous machine learning models. Extensive experiments conducted on the Chinese public-listed corporate rating dataset which we build, prove that CCR-CNN outperforms the state-of-the-art methods consistently.

q-fin.RM↗

arXiv2019

Experimental Observation of High Intrinsic Thermal Conductivity of AlN

AlN is an ultra-wide bandgap semiconductor which has been developed for applications including power electronics and optoelectronics. Thermal management of these applications is the key for stable device performance and allowing for long lifetimes. AlN, with its potentially high thermal conductivity, can play an important role serving as a dielectric layer, growth substrate, and heat spreader to improve device performance. However, the intrinsic high thermal conductivity of bulk AlN predicted by theoretical calculations has not been experimentally observed because of the difficulty in producing materials with low vacancy and impurity levels, and other associated defect complexes in AlN which can decrease the thermal conductivity. This work reports the growth of thick AlN layers by MOCVD with an air-pocketed AlN layer and the first experimental observation of intrinsic thermal conductivity from 130 K to 480 K that matches density-function-theory calculations for single crystal AlN, producing some of the highest values ever measured. Detailed material characterizations confirm the high quality of these AlN samples with one or two orders of magnitude lower impurity concentrations than seen in commercially available bulk AlN. Measurements of these commercially available bulk AlN substrates from 80 K to 480 K demonstrated a lower thermal conductivity, as expected. A theoretical thermal model is built to interpret the measured temperature dependent thermal conductivity. Our work demonstrates that it is possible to obtain theoretically high values of thermal conductivity in AlN and such films may impact the thermal management and reliability of future electronic and optoelectronics devices.

cond-mat.mtrl-sci↗

arXiv2019

Load Forecasting Model and Day-ahead Operation Strategy for City-located EV Quick Charge Stations

Charging demands of electric vehicles (EVs) are sharply increasing due to the rapid development of EVs. Hence, reliable and convenient quick charge stations are required to respond to the needs of EV drivers. Due to the uncertainty of EV charging loads, load forecasting becomes vital for the operation of quick charge stations to formulate the day-ahead plan. In this paper, based on trip chain theory and EV user behaviour, an EV charging load forecasting model is established for quick charge station operators. This model is capable of forecasting the charging demand of a city-located quick charge station during the next day, where the Monte-Carlo simulation method is applied. Furthermore, based on the forecasting model, a day-ahead profit-oriented operation strategy for such stations is derived. The simulation results support the effectiveness of this forecasting model and the operation strategy. The conclusions of this paper are as follows: 1) The charging load forecasting model ensures operators to grasp the feature of the charging load of the next day. 2) The revenue of the quick charge station can be dramatically increased by applying the proposed day-head operation strategy.

eess.SY↗

arXiv2019

Solving Differential Equation with Constrained Multilayer Feedforward Network

In this paper, we present a novel framework to solve differential equations based on multilayer feedforward network. Previous works indicate that solvers based on neural network have low accuracy due to that the boundary conditions are not satisfied accurately. The boundary condition is now inserted directly into the model as boundary term, and the model is a combination of a boundary term and a multilayer feedforward network with its weight function. As the boundary condition becomes predefined constraintion in the model itself, the neural network is trained as an unconstrained optimization problem. This leads to both ease of training and high accuracy. Due to universal convergency of multilayer feedforward networks, the new method is a general approach in solving different types of differential equations. Numerical examples solving ODEs and PDEs with Dirichlet boundary condition are presented and discussed.

math.NA↗

arXiv2017

Magnon and Phonon Dispersion, Lifetime and Thermal Conductivity of Iron from Spin-Lattice Dynamics Simulations

In recent years, the fundamental physics of spin-thermal (i.e., magnon-phonon) interaction has attracted significant experimental and theoretical interests given its potential paradigm-shifting impacts in areas like spin-thermoelectrics, spin-caloritronics and spintronics. Modelling studies of the transport of magnons and phonons in magnetic crystals are very rare. In this paper, we use spin-lattice dynamics (SLD) simulations to model ferromagnetic crystalline iron, where the spin and lattice systems are coupled through the atomic position-dependent exchange function, and thus the interaction between magnon and phonon is naturally considered. We then present a method combining SLD simulations with spectral energy analysis to calculate the magnon and phonon harmonic (e.g., dispersion, specific heat, group velocity) and anharmonic (e.g., scattering rate) properties, based on which their thermal conductivity values are calculated. This work represents an example of using SLD simulations to understand the transport properties involving coupled magnon and phonon dynamics.

cond-mat.mtrl-sci↗