arXiv subjects

Yifan Yang

Publications and source records attributed to Yifan Yang.

At least 325 records · Page 18Linked to original sources

arXiv2021

A Feature Consistency Driven Attention Erasing Network for Fine-Grained Image Retrieval

Large-scale fine-grained image retrieval has two main problems. First, low dimensional feature embedding can fasten the retrieval process but bring accuracy reduce due to overlooking the feature of significant attention regions of images in fine-grained datasets. Second, fine-grained images lead to the same category query hash codes mapping into the different cluster in database hash latent space. To handle these two issues, we propose a feature consistency driven attention erasing network (FCAENet) for fine-grained image retrieval. For the first issue, we propose an adaptive augmentation module in FCAENet, which is selective region erasing module (SREM). SREM makes the network more robust on subtle differences of fine-grained task by adaptively covering some regions of raw images. The feature extractor and hash layer can learn more representative hash code for fine-grained images by SREM. With regard to the second issue, we fully exploit the pair-wise similarity information and add the enhancing space relation loss (ESRL) in FCAENet to make the vulnerable relation stabler between the query hash code and database hash code. We conduct extensive experiments on five fine-grained benchmark datasets (CUB2011, Aircraft, NABirds, VegFru, Food101) for 12bits, 24bits, 32bits, 48bits hash code. The results show that FCAENet achieves the state-of-the-art (SOTA) fine-grained retrieval performance compared with other methods.

cs.CV↗

arXiv2021

PRGC: Potential Relation and Global Correspondence Based Joint Relational Triple Extraction

Joint extraction of entities and relations from unstructured texts is a crucial task in information extraction. Recent methods achieve considerable performance but still suffer from some inherent limitations, such as redundancy of relation prediction, poor generalization of span-based extraction and inefficiency. In this paper, we decompose this task into three subtasks, Relation Judgement, Entity Extraction and Subject-object Alignment from a novel perspective and then propose a joint relational triple extraction framework based on Potential Relation and Global Correspondence (PRGC). Specifically, we design a component to predict potential relations, which constrains the following entity extraction to the predicted relation subset rather than all relations; then a relation-specific sequence tagging component is applied to handle the overlapping problem between subjects and objects; finally, a global correspondence component is designed to align the subject and object into a triple with low-complexity. Extensive experiments show that PRGC achieves state-of-the-art performance on public benchmarks with higher efficiency and delivers consistent performance gain on complex scenarios of overlapping triples.

cs.CL↗

arXiv2021

Modular Forms and $k$-colored Generalized Frobenius Partitions

Let $k$ and $n$ be positive integers. Let $cϕ_{k}(n)$ denote the number of $k$-colored generalized Frobenius partitions of $n$ and $\mathrm{C}Φ_k(q)$ be the generating function of $cϕ_{k}(n)$. In this article, we study $\mathrm{C}Φ_k(q)$ using the theory of modular forms and discover new surprising properties of $\mathrm{C}Φ_k(q)$.

math.NT↗

arXiv2021

SRDiff: Single Image Super-Resolution with Diffusion Probabilistic Models

Single image super-resolution (SISR) aims to reconstruct high-resolution (HR) images from the given low-resolution (LR) ones, which is an ill-posed problem because one LR image corresponds to multiple HR images. Recently, learning-based SISR methods have greatly outperformed traditional ones, while suffering from over-smoothing, mode collapse or large model footprint issues for PSNR-oriented, GAN-driven and flow-based methods respectively. To solve these problems, we propose a novel single image super-resolution diffusion probabilistic model (SRDiff), which is the first diffusion-based model for SISR. SRDiff is optimized with a variant of the variational bound on the data likelihood and can provide diverse and realistic SR predictions by gradually transforming the Gaussian noise into a super-resolution (SR) image conditioned on an LR input through a Markov chain. In addition, we introduce residual prediction to the whole framework to speed up convergence. Our extensive experiments on facial and general benchmarks (CelebA and DIV2K datasets) show that 1) SRDiff can generate diverse SR results in rich details with state-of-the-art performance, given only one LR input; 2) SRDiff is easy to train with a small footprint; and 3) SRDiff can perform flexible image manipulation including latent space interpolation and content fusion.

cs.CV↗

arXiv2021

Quasimodular forms and modular differential equations which are not apparent at cusps: I

In this paper, we explore a two-way connection between quasimodular forms of depth $1$ and a class of second-order modular differential equations with regular singularities on the upper half-plane and the cusps. Here we consider the cases $Γ=Γ_0^+(N)$ generated by $Γ_0(N)$ and the Atkin-Lehner involutions for $N=1,2,3$ ($Γ_0^+(1)=\mathrm{SL}(2,\mathbb Z)$). Firstly, we note that a quasimodular form of depth $1$, after divided by some modular form with the same weight, is a solution of a modular differential equation. Our main results are the converse of the above statement for the groups $Γ_0^+(N)$, $N=1,2,3$.

math.NT↗

arXiv2021

Lifelong Learning based Disease Diagnosis on Clinical Notes

Current deep learning based disease diagnosis systems usually fall short in catastrophic forgetting, i.e., directly fine-tuning the disease diagnosis model on new tasks usually leads to abrupt decay of performance on previous tasks. What is worse, the trained diagnosis system would be fixed once deployed but collecting training data that covers enough diseases is infeasible, which inspires us to develop a lifelong learning diagnosis system. In this work, we propose to adopt attention to combine medical entities and context, embedding episodic memory and consolidation to retain knowledge, such that the learned model is capable of adapting to sequential disease-diagnosis tasks. Moreover, we establish a new benchmark, named Jarvis-40, which contains clinical notes collected from various hospitals. Our experiments show that the proposed method can achieve state-of-the-art performance on the proposed benchmark.

cs.AI↗

arXiv2020

Study of using machine learning for level 1 trigger decision in JUNO experiment

A study on the use of a machine learning algorithm for the level 1 trigger decision in the JUNO experiment ispresented. JUNO is a medium baseline neutrino experiment in construction in China, with the main goal of determining the neutrino mass hierarchy. A large liquid scintillator (LS)volume will detect the electron antineutrinos issued from nuclear reactors. The LS detector is instrumented by around 20000 large photomultiplier tubes. The hit information from each PMT will be collected into a center trigger unit for the level 1 trigger decision. The current trigger algorithm used to select a neutrino signal event is based on a fast vertex reconstruction. We propose to study an alternative level 1 (L1) trigger in order to achieve a similar performance as the vertex fitting trigger but with less logic resources by using firmware implemented machine learning model at the L1 trigger level. We treat the trigger decision as a classification problem and train a Multi-Layer Perceptron (MLP)model to distinguish the signal events with an energy higher than a certain threshold from noise events. We use JUNO software to generate datasets which include 100K physics events with noise and 100K pure noise events coming from PMT dark noise.For events with energy higher than 100 keV, the L1 trigger based on the converged MLP model can achieve an efficiency higher than 99%. After the training performed on simulations,we successfully implemented the trained model into a Kintex 7FPGA. We present the technical details of the neural network development and training, as well as its implementation in the hardware with the FPGA programming. Finally the performance of the L1 trigger MLP implementation is discussed.

physics.ins-det↗

arXiv2020

Automatic test system of the back-end card for the JUNO experiment

The Jiangmen Underground Neutrino Observatory (JUNO) is a medium-baseline neutrino experiment under construction in China, with the goal to determine the neutrino mass hierarchy. The JUNO electronics readout system consists of an underwater front-end electronics system and an outside-water back-end electronics system. These two parts are connected by 100-meter Ethernet cables and power cables. The back-end card (BEC) is the part of the JUNO electronics readout system used to link the underwater boxes to the trigger system is connected to transmit the system clock and triggered signals. Each BEC is connected to 48 underwater boxes, and in total around 150 BECs are needed. It is essential to verify the physical layer links before applying real connection with the underwater system. Therefore, our goal is to build an automatic test system to check the physical link performance. The test system is based on a custom designed FPGA board, in order to make the design general, only JTAG is used as the interface to the PC. The system can generate and check different data pattern at different speeds for 96 channels simultaneously. The test results of 1024 continuously clock cycles are automatically uploaded to PC periodically. We describe the setup of the automatic test system of the BEC and present the latest test results.

physics.ins-det↗

arXiv2020

Modular units and cuspidal divisor classes on $X_0(n^2M)$ with $n|24$ and $M$ squarefree

For a positive integer $N$, let $\mathscr C(N)$ be the subgroup of $J_0(N)$ generated by the equivalence classes of cuspidal divisors of degree $0$ and $\mathscr C(N)(\mathbb Q):=\mathscr C(N)\cap J_0(N)(\mathbb Q)$ be its $\mathbb Q$-rational subgroup. Let also $\mathscr C_{\mathbb Q}(N)$ be the subgroup of $\mathscr C(N)(\mathbb Q)$ generated by $\mathbb Q$-rational cuspidal divisors. We prove that when $N=n^2M$ for some integer $n$ dividing $24$ and some squarefree integer $M$, the two groups $\mathscr C(N)(\mathbb Q)$ and $\mathscr C_{\mathbb Q}(N)$ are equal. To achieve this, we show that all modular units on $X_0(N)$ on such $N$ are products of functions of the form $η(mτ+k/h)$, $mh^2|N$ and $k\in\mathbb Z$ and determine the necessary and sufficient conditions for products of such functions to be modular units on $X_0(N)$.

math.NT↗

arXiv2020

Experiments of Federated Learning for COVID-19 Chest X-ray Images

AI plays an important role in COVID-19 identification. Computer vision and deep learning techniques can assist in determining COVID-19 infection with Chest X-ray Images. However, for the protection and respect of the privacy of patients, the hospital's specific medical-related data did not allow leakage and sharing without permission. Collecting such training data was a major challenge. To a certain extent, this has caused a lack of sufficient data samples when performing deep learning approaches to detect COVID-19. Federated Learning is an available way to address this issue. It can effectively address the issue of data silos and get a shared model without obtaining local data. In the work, we propose the use of federated learning for COVID-19 data training and deploy experiments to verify the effectiveness. And we also compare performances of four popular models (MobileNet, ResNet18, MoblieNet, and COVID-Net) with the federated learning framework and without the framework. This work aims to inspire more researches on federated learning about COVID-19.

eess.IV↗

arXiv2020

COVID-19 Public Opinion and Emotion Monitoring System Based on Time Series Thermal New Word Mining

With the spread and development of new epidemics, it is of great reference value to identify the changing trends of epidemics in public emotions. We designed and implemented the COVID-19 public opinion monitoring system based on time series thermal new word mining. A new word structure discovery scheme based on the timing explosion of network topics and a Chinese sentiment analysis method for the COVID-19 public opinion environment is proposed. Establish a "Scrapy-Redis-Bloomfilter" distributed crawler framework to collect data. The system can judge the positive and negative emotions of the reviewer based on the comments, and can also reflect the depth of the seven emotions such as Hopeful, Happy, and Depressed. Finally, we improved the sentiment discriminant model of this system and compared the sentiment discriminant error of COVID-19 related comments with the Jiagu deep learning model. The results show that our model has better generalization ability and smaller discriminant error. We designed a large data visualization screen, which can clearly show the trend of public emotions, the proportion of various emotion categories, keywords, hot topics, etc., and fully and intuitively reflect the development of public opinion.

cs.IR↗

arXiv2020

Ramanujan-type $1/π$-series from bimodular forms

We develop an approach to establish $1/π$-series from bimodular forms. Utilizing this approach, we obtain new families of $2$-variable $1/π$-series associated to Zagier's sporadic Apery-like sequences.

math.NT↗

arXiv2020

Synetgy: Algorithm-hardware Co-design for ConvNet Accelerators on Embedded FPGAs

Using FPGAs to accelerate ConvNets has attracted significant attention in recent years. However, FPGA accelerator design has not leveraged the latest progress of ConvNets. As a result, the key application characteristics such as frames-per-second (FPS) are ignored in favor of simply counting GOPs, and results on accuracy, which is critical to application success, are often not even reported. In this work, we adopt an algorithm-hardware co-design approach to develop a ConvNet accelerator called Synetgy and a novel ConvNet model called DiracDeltaNet$^{\dagger}$. Both the accelerator and ConvNet are tailored to FPGA requirements. DiracDeltaNet, as the name suggests, is a ConvNet with only $1\times 1$ convolutions while spatial convolutions are replaced by more efficient shift operations. DiracDeltaNet achieves competitive accuracy on ImageNet (88.7\% top-5), but with 42$\times$ fewer parameters and 48$\times$ fewer OPs than VGG16. We further quantize DiracDeltaNet's weights to 4-bit and activations to 4-bits, with less than 1\% accuracy loss. These quantizations exploit well the nature of FPGA hardware. In short, DiracDeltaNet's small model size, low computational OP count, low precision and simplified operators allow us to co-design a highly customized computing unit for an FPGA. We implement the computing units for DiracDeltaNet on an Ultra96 SoC system through high-level synthesis. Our accelerator's final top-5 accuracy of 88.1\% on ImageNet, is higher than all the previously reported embedded FPGA accelerators. In addition, the accelerator reaches an inference speed of 66.3 FPS on the ImageNet classification task, surpassing prior works with similar accuracy by at least 11.6$\times$.

cs.CV↗

arXiv2020

The smallest parts function associated with $ω(q)$

We establish two families of congruences modulo powers of 5 for the Fourier coefficients of $(2E_2(2τ)-E_2(τ))η(2τ)^{-1}$, where $E_2(τ)$ is the weight 2 Eisenstein series and $η(τ)$ is the Dedekind eta function. This allows us to prove similar congruences for two smallest parts functions. The first function is $\mathrm{spt}_ω(n)$, which was introduced by Andrews, Dixit and Yee and associated with Ramanujan/Watson's third order mock theta function $ω(q)$. The second one is $\mathrm{spt}_{C5}(n)$, which appeared in the work of Garvan and Jennings-Shaffer. Moreover, we confirm two conjectural congruences of Wang.

math.NT↗

arXiv2020

Moments of ranks and cranks, and Quotients of Eisenstein Series and the Dedekind Eta Function

Atkin and Garvan introduced the functions $N_k(n)$ and $M_k(n)$, which denote the $k$-th moments of ranks and cranks in the theory of partitions. Let $e_{2r}(n)$ be the $n$-th Fourier coefficient of $E_{2r}(τ)/η(τ)$, where $E_{2r}(τ)$ is the classical Eisenstein series of weight $2r$ and $η(τ)$ is the Dedekind eta function. Via the theory of quasi-modular forms, we find that for $k \leq 5$, $N_k(n)$ and $M_k(n)$ can be expressed using $e_{2r}(n)$ ($2\leq r \leq k$), $p(n)$ and $N_2(n)$. For $k>5$, additional functions are required for such expressions. For $r\in \{2, 3, 4, 5, 7\}$, by studying the action of Hecke operators on $E_{2r}(τ)/η(τ)$, we provide explicit congruences modulo arbitrary powers of primes for $e_{2r}(n)$. Moreover, for $\ell \in \{5, 7, 11, 13\}$ and any $k\geq 1$, we present uniform methods for finding nice representations for $\sum_{n=0}^\infty e_{2r}\left(\frac{\ell^{k}n+1}{24}\right)q^n$, which work for every $r\geq 2$. These representations allow us to prove congruences modulo powers of $\ell$, and we have done so for $e_4(n)$ and $e_6(n)$ as examples. Based on the congruences satisfied by $e_{2r}(n)$, we establish congruences modulo arbitrary powers of $\ell$ for the moments and symmetrized moments of ranks and cranks as well as higher order $\mathrm{spt}$-functions.

math.NT↗

arXiv2020

A New Unified Deep Learning Approach with Decomposition-Reconstruction-Ensemble Framework for Time Series Forecasting

A new variational mode decomposition (VMD) based deep learning approach is proposed in this paper for time series forecasting problem. Firstly, VMD is adopted to decompose the original time series into several sub-signals. Then, a convolutional neural network (CNN) is applied to learn the reconstruction patterns on the decomposed sub-signals to obtain several reconstructed sub-signals. Finally, a long short term memory (LSTM) network is employed to forecast the time series with the decomposed sub-signals and the reconstructed sub-signals as inputs. The proposed VMD-CNN-LSTM approach is originated from the decomposition-reconstruction-ensemble framework, and innovated by embedding the reconstruction, single forecasting, and ensemble steps in a unified deep learning approach. To verify the forecasting performance of the proposed approach, four typical time series datasets are introduced for empirical analysis. The empirical results demonstrate that the proposed approach outperforms consistently the benchmark approaches in terms of forecasting accuracy, and also indicate that the reconstructed sub-signals obtained by CNN is of importance for further improving the forecasting performance.

stat.ML↗

arXiv2020

Water affects morphogenesis of growing aquatic plant leaves

Lotus leaves floating on water usually experience short-wavelength edge wrinkling that decays toward the center, while the leaves growing above water normally morph into a global bending cone shape with long rippled waves near the edge. Observations suggest that the underlying water (liquid substrate) significantly affects the morphogenesis of leaves. To understand the biophysical mechanism under such phenomena, we develop mathematical models that can effectively account for inhomogeneous differential growth of floating and free-standing leaves, to quantitatively predict formation and evolution of their morphology. We find, both theoretically and experimentally, that the short-wavelength buckled configuration is energetically favorable for growing membranes lying on liquid, while the global buckling shape is more preferable for suspended ones. Other influencing factors such as stem/vein, heterogeneity and dimension are also investigated. Our results provide a fundamental insight into a variety of plant morphogenesis affected by water foundation and suggest that such surface instabilities can be harnessed for morphology control of biomimetic deployable structures using substrate or edge actuation.

cond-mat.soft↗

arXiv2019

GRID: a Student Project to Monitor the Transient Gamma-Ray Sky in the Multi-Messenger Astronomy Era

The Gamma-Ray Integrated Detectors (GRID) is a space mission concept dedicated to monitoring the transient gamma-ray sky in the energy range from 10 keV to 2 MeV using scintillation detectors onboard CubeSats in low Earth orbits. The primary targets of GRID are the gamma-ray bursts (GRBs) in the local universe. The scientific goal of GRID is, in synergy with ground-based gravitational wave (GW) detectors such as LIGO and VIRGO, to accumulate a sample of GRBs associated with the merger of two compact stars and study jets and related physics of those objects. It also involves observing and studying other gamma-ray transients such as long GRBs, soft gamma-ray repeaters, terrestrial gamma-ray flashes, and solar flares. With multiple CubeSats in various orbits, GRID is unaffected by the Earth occultation and serves as a full-time and all-sky monitor. Assuming a horizon of 200 Mpc for ground-based GW detectors, we expect to see a few associated GW-GRB events per year. With about 10 CubeSats in operation, GRID is capable of localizing a faint GRB like 170817A with a 90% error radius of about 10 degrees, through triangulation and flux modulation. GRID is proposed and developed by students, with considerable contribution from undergraduate students, and will remain operated as a student project in the future. The current GRID collaboration involves more than 20 institutes and keeps growing. On August 29th, the first GRID detector onboard a CubeSat was launched into a Sun-synchronous orbit and is currently under test.

astro-ph.IM↗