arXiv subjects

Zhiyuan Yan

Publications and source records attributed to Zhiyuan Yan.

At least 73 records · Page 4Linked to original sources

arXiv2023

UCF: Uncovering Common Features for Generalizable Deepfake Detection

Deepfake detection remains a challenging task due to the difficulty of generalizing to new types of forgeries. This problem primarily stems from the overfitting of existing detection methods to forgery-irrelevant features and method-specific patterns. The latter has been rarely studied and not well addressed by previous works. This paper presents a novel approach to address the two types of overfitting issues by uncovering common forgery features. Specifically, we first propose a disentanglement framework that decomposes image information into three distinct components: forgery-irrelevant, method-specific forgery, and common forgery features. To ensure the decoupling of method-specific and common forgery features, a multi-task learning strategy is employed, including a multi-class classification that predicts the category of the forgery method and a binary classification that distinguishes the real from the fake. Additionally, a conditional decoder is designed to utilize forgery features as a condition along with forgery-irrelevant features to generate reconstructed images. Furthermore, a contrastive regularization technique is proposed to encourage the disentanglement of the common and specific forgery features. Ultimately, we only utilize the common forgery features for the purpose of generalizable deepfake detection. Extensive evaluations demonstrate that our framework can perform superior generalization than current state-of-the-art methods.

cs.CV↗

arXiv2023

DeepfakeBench: A Comprehensive Benchmark of Deepfake Detection

A critical yet frequently overlooked challenge in the field of deepfake detection is the lack of a standardized, unified, comprehensive benchmark. This issue leads to unfair performance comparisons and potentially misleading results. Specifically, there is a lack of uniformity in data processing pipelines, resulting in inconsistent data inputs for detection models. Additionally, there are noticeable differences in experimental settings, and evaluation strategies and metrics lack standardization. To fill this gap, we present the first comprehensive benchmark for deepfake detection, called DeepfakeBench, which offers three key contributions: 1) a unified data management system to ensure consistent input across all detectors, 2) an integrated framework for state-of-the-art methods implementation, and 3) standardized evaluation metrics and protocols to promote transparency and reproducibility. Featuring an extensible, modular-based codebase, DeepfakeBench contains 15 state-of-the-art detection methods, 9 deepfake datasets, a series of deepfake detection evaluation protocols and analysis tools, as well as comprehensive evaluations. Moreover, we provide new insights based on extensive analysis of these evaluations from various perspectives (e.g., data augmentations, backbones). We hope that our efforts could facilitate future research and foster innovation in this increasingly critical domain. All codes, evaluations, and analyses of our benchmark are publicly available at https://github.com/SCLBD/DeepfakeBench.

cs.CV↗

arXiv2023

Not Just Learning from Others but Relying on Yourself: A New Perspective on Few-Shot Segmentation in Remote Sensing

Few-shot segmentation (FSS) is proposed to segment unknown class targets with just a few annotated samples. Most current FSS methods follow the paradigm of mining the semantics from the support images to guide the query image segmentation. However, such a pattern of `learning from others' struggles to handle the extreme intra-class variation, preventing FSS from being directly generalized to remote sensing scenes. To bridge the gap of intra-class variance, we develop a Dual-Mining network named DMNet for cross-image mining and self-mining, meaning that it no longer focuses solely on support images but pays more attention to the query image itself. Specifically, we propose a Class-public Region Mining (CPRM) module to effectively suppress irrelevant feature pollution by capturing the common semantics between the support-query image pair. The Class-specific Region Mining (CSRM) module is then proposed to continuously mine the class-specific semantics of the query image itself in a `filtering' and `purifying' manner. In addition, to prevent the co-existence of multiple classes in remote sensing scenes from exacerbating the collapse of FSS generalization, we also propose a new Known-class Meta Suppressor (KMS) module to suppress the activation of known-class objects in the sample. Extensive experiments on the iSAID and LoveDA remote sensing datasets have demonstrated that our method sets the state-of-the-art with a minimum number of model parameters. Significantly, our model with the backbone of Resnet-50 achieves the mIoU of 49.58% and 51.34% on iSAID under 1-shot and 5-shot settings, outperforming the state-of-the-art method by 1.8% and 1.12%, respectively. The code is publicly available at https://github.com/HanboBizl/DMNet.

cs.CV↗

arXiv2023

iEDA: An Open-Source Intelligent Physical Implementation Toolkit and Library

Open-source EDA shows promising potential in unleashing EDA innovation and lowering the cost of chip design. This paper presents an open-source EDA project, iEDA, aiming for building a basic infrastructure for EDA technology evolution and closing the industrial-academic gap in the EDA area. iEDA now covers the whole flow of physical design (including Floorplan, Placement, CTS, Routing, Timing Optimization etc.), and part of the analysis tools (Static Timing Analysis and Power Analysis). To demonstrate the effectiveness of iEDA, we implement and tape out three chips of different scales (from 700k to 1.5M gates) on different process nodes (110nm and 28nm) with iEDA. iEDA is publicly available from the project home page http://ieda.oscc.cc.

cs.AR↗

arXiv2023

Multimodal Graph Learning for Deepfake Detection

Existing deepfake detectors face several challenges in achieving robustness and generalization. One of the primary reasons is their limited ability to extract relevant information from forgery videos, especially in the presence of various artifacts such as spatial, frequency, temporal, and landmark mismatches. Current detectors rely on pixel-level features that are easily affected by unknown disturbances or facial landmarks that do not provide sufficient information. Furthermore, most detectors cannot utilize information from multiple domains for detection, leading to limited effectiveness in identifying deepfake videos. To address these limitations, we propose a novel framework, namely Multimodal Graph Learning (MGL) that leverages information from multiple modalities using two GNNs and several multimodal fusion modules. At the frame level, we employ a bi-directional cross-modal transformer and an adaptive gating mechanism to combine the features from the spatial and frequency domains with the geometric-enhanced landmark features captured by a GNN. At the video level, we use a Graph Attention Network (GAT) to represent each frame in a video as a node in a graph and encode temporal information into the edges of the graph to extract temporal inconsistency between frames. Our proposed method aims to effectively identify and utilize distinguishing features for deepfake detection. We evaluate the effectiveness of our method through extensive experiments on widely-used benchmarks and demonstrate that our method outperforms the state-of-the-art detectors in terms of generalization ability and robustness against unknown disturbances.

cs.CV↗

arXiv2023

Addressing Variable Dependency in GNN-based SAT Solving

Boolean satisfiability problem (SAT) is fundamental to many applications. Existing works have used graph neural networks (GNNs) for (approximate) SAT solving. Typical GNN-based end-to-end SAT solvers predict SAT solutions concurrently. We show that for a group of symmetric SAT problems, the concurrent prediction is guaranteed to produce a wrong answer because it neglects the dependency among Boolean variables in SAT problems. % We propose AsymSAT, a GNN-based architecture which integrates recurrent neural networks to generate dependent predictions for variable assignments. The experiment results show that dependent variable prediction extends the solving capability of the GNN-based method as it improves the number of solved SAT instances on large test sets.

cs.AI↗

arXiv2022

Full Geometric Control of Hidden Color Information in Diffraction Gratings under Angled White Light Illumination

Under white light illumination, gratings produce an angular distribution of wavelengths dependent on the diffraction order and geometric parameters. However, previous studies of gratings are limited to at least one geometric parameter (height, periodicity, orientation, angle of incidence) kept constant. Here, we vary all geometric parameters in the gratings using a versatile nanofabrication technique, two-photon polymerization lithography, to encode hidden color information through 2 design approaches. The first approach hides color information by decoupling the effects of grating height and periodicity under normal and oblique incidence. The second approach hides multiple sets of color information by arranging gratings in sectors around semi-circular pixels. Different images are revealed with negligible crosstalk under oblique incidence and varying sample rotation angles. Our analysis shows that an angular separation >= 10° between adjacent sectors is required to suppress crosstalk. This work has potential applications in information storage and security watermarks.

physics.optics↗

arXiv2022

HelixADMET: a robust and endpoint extensible ADMET system incorporating self-supervised knowledge transfer

Accurate ADMET (an abbreviation for "absorption, distribution, metabolism, excretion, and toxicity") predictions can efficiently screen out undesirable drug candidates in the early stage of drug discovery. In recent years, multiple comprehensive ADMET systems that adopt advanced machine learning models have been developed, providing services to estimate multiple endpoints. However, those ADMET systems usually suffer from weak extrapolation ability. First, due to the lack of labelled data for each endpoint, typical machine learning models perform frail for the molecules with unobserved scaffolds. Second, most systems only provide fixed built-in endpoints and cannot be customised to satisfy various research requirements. To this end, we develop a robust and endpoint extensible ADMET system, HelixADMET (H-ADMET). H-ADMET incorporates the concept of self-supervised learning to produce a robust pre-trained model. The model is then fine-tuned with a multi-task and multi-stage framework to transfer knowledge between ADMET endpoints, auxiliary tasks, and self-supervised tasks. Our results demonstrate that H-ADMET achieves an overall improvement of 4%, compared with existing ADMET systems on comparable endpoints. Additionally, the pre-trained model provided by H-ADMET can be fine-tuned to generate new and customised ADMET endpoints, meeting various demands of drug research and development requirements.

q-bio.BM↗

arXiv2021

FAIR1M: A Benchmark Dataset for Fine-grained Object Recognition in High-Resolution Remote Sensing Imagery

With the rapid development of deep learning, many deep learning-based approaches have made great achievements in object detection task. It is generally known that deep learning is a data-driven method. Data directly impact the performance of object detectors to some extent. Although existing datasets have included common objects in remote sensing images, they still have some limitations in terms of scale, categories, and images. Therefore, there is a strong requirement for establishing a large-scale benchmark on object detection in high-resolution remote sensing images. In this paper, we propose a novel benchmark dataset with more than 1 million instances and more than 15,000 images for Fine-grAined object recognItion in high-Resolution remote sensing imagery which is named as FAIR1M. All objects in the FAIR1M dataset are annotated with respect to 5 categories and 37 sub-categories by oriented bounding boxes. Compared with existing detection datasets dedicated to object detection, the FAIR1M dataset has 4 particular characteristics: (1) it is much larger than other existing object detection datasets both in terms of the quantity of instances and the quantity of images, (2) it provides more rich fine-grained category information for objects in remote sensing images, (3) it contains geographic information such as latitude, longitude and resolution, (4) it provides better image quality owing to a careful data cleaning procedure. To establish a baseline for fine-grained object recognition, we propose a novel evaluation method and benchmark fine-grained object detection tasks and a visual classification task using several State-Of-The-Art (SOTA) deep learning-based models on our FAIR1M dataset. Experimental results strongly indicate that the FAIR1M dataset is closer to practical application and it is considerably more challenging than existing datasets.

cs.CV↗

arXiv2021

A Novel Key Generation Scheme Using Quaternary PUF Responses and Wiretap Polar Coding

Physical unclonable functions (PUFs) are widely considered in secret key generation for resource constrained devices. However, PUFs require additional hardware overhead. In this paper, we focus on developing a PUF-efficient, robust, and secure key generation scheme. First, a novel method for extracting quaternary PUF responses is proposed to increase the entropy of a PUF response, in which a 2-bit response is extracted from evaluating a single PUF cell multiple times. The probability masses of the responses can be adjusted by setting parameters appropriately. Then, a chosen secret model based fuzzy extractor (FE) is designed to extract secret keys from the quaternary PUF responses. To improve the security of this FE, it is modeled as a wiretap channel system, and wiretap polar coding is adopted to reduce secrecy leakage. An upper bound of secrecy leakage is also given in this paper, and it suggests that an arbitrarily small (even zero) leakage can be achieved by properly choosing parameters of the quaternary PUF responses generation. Comparison results show that the required number of PUF cells to achieve the same level of secrecy in our scheme is as low as half that of the state-of-the-art schemes.

cs.CR↗

arXiv2019

Inference with Hybrid Bio-hardware Neural Networks

To understand the learning process in brains, biologically plausible algorithms have been explored by modeling the detailed neuron properties and dynamics. On the other hand, simplified multi-layer models of neural networks have shown great success on computational tasks such as image classification and speech recognition. However, the computational models that can achieve good accuracy for these learning applications are very different from the bio-plausible models. This paper studies whether a bio-plausible model of a in vitro living neural network can be used to perform machine learning tasks and achieve good inference accuracy. A novel two-layer bio-hardware hybrid neural network is proposed. The biological layer faithfully models variations of synapses, neurons, and network sparsity in in vitro living neural networks. The hardware layer is a computational fully-connected layer that tunes parameters to optimize for accuracy. Several techniques are proposed to improve the inference accuracy of the proposed hybrid neural network. For instance, an adaptive pre-processing technique helps the proposed neural network to achieve good learning accuracy for different living neural network sparsity. The proposed hybrid neural network with realistic neuron parameters and variations achieves a 98.3% testing accuracy for the handwritten digit recognition task on the full MNIST dataset.

cs.NE↗

arXiv2018

Avalon: Building an Operating System for Robotcenter

This paper envisions a scenario that hundreds of heterogeneous robots form a robotcenter which can be shared by multiple users and used like a single powerful robot to perform complex tasks. However, current multi-robot systems are either unable to manage heterogeneous robots or unable to support multiple concurrent users. Inspired by the design of modern datacenter OSes, we propose Avalon, a robot operating system with two-level scheduling scheme which is widely adopted in datacenters for Internet services and cloud computing. Specifically, Avalon integrates three important features together: (1) Instead of allocating a whole robot, Avalon classifies fine-grained robot resources into three categories to distinguish which fine-grained resources can be shared by multi-robot frameworks simultaneously. (2) Avalon adopts a location based resource allocation policy to substantially reduce scheduling overhead. (3) Avalon enables robots to offload computation intensive tasks to the clouds.We have implemented and evaluated Avalon on robots on both simulated environments and real world.

cs.RO↗

arXiv2018

A Supervised STDP-based Training Algorithm for Living Neural Networks

Neural networks have shown great potential in many applications like speech recognition, drug discovery, image classification, and object detection. Neural network models are inspired by biological neural networks, but they are optimized to perform machine learning tasks on digital computers. The proposed work explores the possibilities of using living neural networks in vitro as basic computational elements for machine learning applications. A new supervised STDP-based learning algorithm is proposed in this work, which considers neuron engineering constrains. A 74.7% accuracy is achieved on the MNIST benchmark for handwritten digit recognition.

cs.NE↗

arXiv2016

Local Codes with Cooperative Repair in Distributed Storage System

Recently, the research on local repair codes is mainly confined to repair the failed nodes within each repair group. But if the extreme cases occur that the entire repair group has failed, the local code stored in the failed group need to be recovered as a whole. In this paper, local codes with cooperative repair, in which the local codes are constructed based on minimum storage regeneration (MSR) codes, is proposed to achieve repairing the failed groups. Specifically, the proposed local codes with cooperative repair construct a kind of mutual interleaving structure among the parity symbols, that the parity symbols of each local code, named as distributed local parity, can be generated by the parity symbols of the MSR codes in its two adjacent local codes. Taking advantage of the structure given, the failed local groups can be repaired cooperatively by their adjacent local groups with lower repair locality, and meanwhile the minimum distance of local codes with cooperative repair is derived. Theoretical analysis and simulation experiments show that, compared with codes with local regeneration (such as MSR-local codes and MBR-local codes), the proposed local codes with cooperative repair have benefits in bandwidth overhead and repair locality for the case of local groups failure.

cs.IT↗

arXiv2015

A multi-mode area-efficient SCL polar decoder

Polar codes are of great interest since they are the first provably capacity-achieving forward error correction codes. To improve throughput and to reduce decoding latency of polar decoders, maximum likelihood (ML) decoding units are used by successive cancellation list (SCL) decoders as well as successive cancellation (SC) decoders. This paper proposes an approximate ML (AML) decoding unit for SCL decoders first. In particular, we investigate the distribution of frozen bits of polar codes designed for both the binary erasure and additive white Gaussian noise channels, and take advantage of the distribution to reduce the complexity of the AML decoding unit, improving the area efficiency of SCL decoders. Furthermore, a multi-mode SCL decoder with variable list sizes and parallelism is proposed. If high throughput or small latency is required, the decoder decodes multiple received codewords in parallel with a small list size. However, if error performance is of higher priority, the multi-mode decoder switches to a serial mode with a bigger list size. Therefore, the multi-mode SCL decoder provides a flexible tradeoff between latency, throughput and error performance, and adapts to different throughput and latency requirements at the expense of small overhead. Hardware implementation and synthesis results show that our polar decoders not only have a better area efficiency but also easily adapt to different communication channels and applications.

cs.IT↗

arXiv2015

Reduced Complexity Belief Propagation Decoders for Polar Codes

Polar codes are newly discovered capacity-achieving codes, which have attracted lots of research efforts. Polar codes can be efficiently decoded by the low-complexity successive cancelation (SC) algorithm and the SC list (SCL) decoding algorithm. The belief propagation (BP) decoding algorithm not only is an alternative to the SC and SCL decoders, but also provides soft outputs that are necessary for joint detection and decoding. Both the BP decoder and the soft cancelation (SCAN) decoder were proposed for polar codes to output soft information about the coded bits. In this paper, first a belief propagation decoding algorithm, called reduced complexity soft cancelation (RCSC) decoding algorithm, is proposed. Let $N$ denote the block length. Our RCSC decoding algorithm needs to store only $5N-3$ log-likelihood ratios (LLRs), significantly less than $4N-2+\frac{N\log_2N}{2}$ and $N(\log_2N+1)$ LLRs needed by the BP and SCAN decoders, respectively, when $N\geqslant 64$. Besides, compared to the SCAN decoding algorithm, our RCSC decoding algorithm eliminates unnecessary additions over the real field. Then the simplified SC (SSC) principle is applied to our RCSC decoding algorithm, and the resulting SSC-aided RCSC (S-RCSC) decoding algorithm further reduces the computational complexity. Finally, based on the S-RCSC decoding algorithm, we propose a corresponding memory efficient decoder architecture, which has better error performance than existing architectures. Besides, our decoder architecture consumes less energy on updating LLRs.

cs.IT↗

arXiv2015

A High Throughput List Decoder Architecture for Polar Codes

While long polar codes can achieve the capacity of arbitrary binary-input discrete memoryless channels when decoded by a low complexity successive cancelation (SC) algorithm, the error performance of the SC algorithm is inferior for polar codes with finite block lengths. The cyclic redundancy check (CRC) aided successive cancelation list (SCL) decoding algorithm has better error performance than the SC algorithm. However, current CRC aided SCL (CA-SCL) decoders still suffer from long decoding latency and limited throughput. In this paper, a reduced latency list decoding (RLLD) algorithm for polar codes is proposed. Our RLLD algorithm performs the list decoding on a binary tree, whose leaves correspond to the bits of a polar code. In existing SCL decoding algorithms, all the nodes in the tree are traversed and all possibilities of the information bits are considered. Instead, our RLLD algorithm visits much fewer nodes in the tree and considers fewer possibilities of the information bits. When configured properly, our RLLD algorithm significantly reduces the decoding latency and hence improves throughput, while introducing little performance degradation. Based on our RLLD algorithm, we also propose a high throughput list decoder architecture, which is suitable for larger block lengths due to its scalable partial sum computation unit. Our decoder architecture has been implemented for different block lengths and list sizes using the TSMC 90nm CMOS technology. The implementation results demonstrate that our decoders achieve significant latency reduction and area efficiency improvement compared with other list polar decoders in the literature.

cs.AR↗

arXiv2015

A hybrid partial sum computation unit architecture for list decoders of polar codes

Although the successive cancelation (SC) algorithm works well for very long polar codes, its error performance for shorter polar codes is much worse. Several SC based list decoding algorithms have been proposed to improve the error performances of both long and short polar codes. A significant step of SC based list decoding algorithms is the updating of partial sums for all decoding paths. In this paper, we first proposed a lazy copy partial sum computation algorithm for SC based list decoding algorithms. Instead of copying partial sums directly, our lazy copy algorithm copies indices of partial sums. Based on our lazy copy algorithm, we propose a hybrid partial sum computation unit architecture, which employs both registers and memories so that the overall area efficiency is improved. Compared with a recent partial sum computation unit for list decoders, when the list size $L=4$, our partial sum computation unit achieves an area saving of 23\% and 63\% for block length $2^{13}$ and $2^{15}$, respectively.

cs.IT↗