arXiv subjects

Zheng Li

Publications and source records attributed to Zheng Li.

At least 325 records · Page 18Linked to original sources

arXiv2023

Entangled X-ray Photon Pair Generation by Free Electron Lasers

Einstein, Podolsky and Rosen's prediction on incompleteness of quantum mechanics was overturned by experimental tests on Bell's inequality that confirmed the existence of quantum entanglement. In X-ray optics, entangled photon pairs can be generated by X-ray parametric down conversion (XPDC), which has certain wavelength window. Meanwhile, free electron laser (FEL) has successfully lased at X-ray frequencies recently. However, FEL is usually seen as a classical light source, and its quantum effects are considered minor corrections to the classical theory. Here we investigate entangled X-ray photon pair emissions in FEL. We establish a theory for coherently amplified entangled photon pair emission from microbunched electron pulses in the undulator. We numerically demonstrate the properties of entangled emission, and provide a scheme to generate highly entangled X-ray photon pairs, which is of great importance in X-ray quantum optics. Our work shows a unique advantage of FELs over synchrotrons in entangled X-ray photon pair generation.

quant-ph↗

arXiv2023

HomoDistil: Homotopic Task-Agnostic Distillation of Pre-trained Transformers

Knowledge distillation has been shown to be a powerful model compression approach to facilitate the deployment of pre-trained language models in practice. This paper focuses on task-agnostic distillation. It produces a compact pre-trained model that can be easily fine-tuned on various tasks with small computational costs and memory footprints. Despite the practical benefits, task-agnostic distillation is challenging. Since the teacher model has a significantly larger capacity and stronger representation power than the student model, it is very difficult for the student to produce predictions that match the teacher's over a massive amount of open-domain training data. Such a large prediction discrepancy often diminishes the benefits of knowledge distillation. To address this challenge, we propose Homotopic Distillation (HomoDistil), a novel task-agnostic distillation approach equipped with iterative pruning. Specifically, we initialize the student model from the teacher model, and iteratively prune the student's neurons until the target width is reached. Such an approach maintains a small discrepancy between the teacher's and student's predictions throughout the distillation process, which ensures the effectiveness of knowledge transfer. Extensive experiments demonstrate that HomoDistil achieves significant improvements on existing baselines.

cs.CL↗

arXiv2023

FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer

Network quantization significantly reduces model inference complexity and has been widely used in real-world deployments. However, most existing quantization methods have been developed mainly on Convolutional Neural Networks (CNNs), and suffer severe degradation when applied to fully quantized vision transformers. In this work, we demonstrate that many of these difficulties arise because of serious inter-channel variation in LayerNorm inputs, and present, Power-of-Two Factor (PTF), a systematic method to reduce the performance degradation and inference complexity of fully quantized vision transformers. In addition, observing an extreme non-uniform distribution in attention maps, we propose Log-Int-Softmax (LIS) to sustain that and simplify inference by using 4-bit quantization and the BitShift operator. Comprehensive experiments on various transformer-based architectures and benchmarks show that our Fully Quantized Vision Transformer (FQ-ViT) outperforms previous works while even using lower bit-width on attention maps. For instance, we reach 84.89% top-1 accuracy with ViT-L on ImageNet and 50.8 mAP with Cascade Mask R-CNN (Swin-S) on COCO. To our knowledge, we are the first to achieve lossless accuracy degradation (~1%) on fully quantized vision transformers. The code is available at https://github.com/megvii-research/FQ-ViT.

cs.CV↗

arXiv2023

Structured air lasing of N2+

Structured light has attracted great interest in scientific and technical fields. Here, we demonstrate the first generation of structured air lasing in N2+ driven by 800 nm femtosecond laser pulses. By focusing a vortex pump beam at 800 nm in N2 gas, we generate a vortex superfluorescent radiation of N2+ at 391 nm, which carries the same photon orbital angular momentum as the pump beam. With the injection of a Gaussian seed beam at 391 nm, the coherent radiation is amplified, but the vorticity is unchanged. A new physical mechanism is revealed in the vortex N2+ superfluorescent radiation: the vortex pump beam transfers the spatial spiral phase into the N2+ gain medium, and the Gaussian seed beam picks up the spatial spiral phase and is then amplified into a vortex beam. Moreover, when we employ a pump beam with a cylindrical vector mode, the Gaussian seed beam is correspondingly amplified into a cylindrical vector beam. Surprisingly, the spatial polarization state of the amplified radiation is identical to that of the vector pump beam regardless of whether the Gaussian seed beam is linearly, elliptically, or circularly polarized. Solving three-dimensional coupled wave equations, we show how a Gaussian beam becomes a cylindrical vector beam in a cylindrically symmetric gain medium. This study provides a novel approach to generating structured light via N2+ air lasing.

physics.optics↗

arXiv2023

DE-FAKE: Detection and Attribution of Fake Images Generated by Text-to-Image Generation Models

Text-to-image generation models that generate images based on prompt descriptions have attracted an increasing amount of attention during the past few months. Despite their encouraging performance, these models raise concerns about the misuse of their generated fake images. To tackle this problem, we pioneer a systematic study on the detection and attribution of fake images generated by text-to-image generation models. Concretely, we first build a machine learning classifier to detect the fake images generated by various text-to-image generation models. We then attribute these fake images to their source models, such that model owners can be held responsible for their models' misuse. We further investigate how prompts that generate fake images affect detection and attribution. We conduct extensive experiments on four popular text-to-image generation models, including DALL$\cdot$E 2, Stable Diffusion, GLIDE, and Latent Diffusion, and two benchmark prompt-image datasets. Empirical results show that (1) fake images generated by various models can be distinguished from real ones, as there exists a common artifact shared by fake images from different models; (2) fake images can be effectively attributed to their source models, as different models leave unique fingerprints in their generated images; (3) prompts with the ``person'' topic or a length between 25 and 75 enable models to generate fake images with higher authenticity. All findings contribute to the community's insight into the threats caused by text-to-image generation models. We appeal to the community's consideration of the counterpart solutions, like ours, against the rapidly-evolving fake image generation.

cs.CR↗

arXiv2023

Backdoor Attacks Against Dataset Distillation

Dataset distillation has emerged as a prominent technique to improve data efficiency when training machine learning models. It encapsulates the knowledge from a large dataset into a smaller synthetic dataset. A model trained on this smaller distilled dataset can attain comparable performance to a model trained on the original training dataset. However, the existing dataset distillation techniques mainly aim at achieving the best trade-off between resource usage efficiency and model utility. The security risks stemming from them have not been explored. This study performs the first backdoor attack against the models trained on the data distilled by dataset distillation models in the image domain. Concretely, we inject triggers into the synthetic data during the distillation procedure rather than during the model training stage, where all previous attacks are performed. We propose two types of backdoor attacks, namely NAIVEATTACK and DOORPING. NAIVEATTACK simply adds triggers to the raw data at the initial distillation phase, while DOORPING iteratively updates the triggers during the entire distillation procedure. We conduct extensive evaluations on multiple datasets, architectures, and dataset distillation techniques. Empirical evaluation shows that NAIVEATTACK achieves decent attack success rate (ASR) scores in some cases, while DOORPING reaches higher ASR scores (close to 1.0) in all cases. Furthermore, we conduct a comprehensive ablation study to analyze the factors that may affect the attack performance. Finally, we evaluate multiple defense mechanisms against our backdoor attacks and show that our attacks can practically circumvent these defense mechanisms.

cs.CR↗

arXiv2022

Curriculum Temperature for Knowledge Distillation

Most existing distillation methods ignore the flexible role of the temperature in the loss function and fix it as a hyper-parameter that can be decided by an inefficient grid search. In general, the temperature controls the discrepancy between two distributions and can faithfully determine the difficulty level of the distillation task. Keeping a constant temperature, i.e., a fixed level of task difficulty, is usually sub-optimal for a growing student during its progressive learning stages. In this paper, we propose a simple curriculum-based technique, termed Curriculum Temperature for Knowledge Distillation (CTKD), which controls the task difficulty level during the student's learning career through a dynamic and learnable temperature. Specifically, following an easy-to-hard curriculum, we gradually increase the distillation loss w.r.t. the temperature, leading to increased distillation difficulty in an adversarial manner. As an easy-to-use plug-in technique, CTKD can be seamlessly integrated into existing knowledge distillation frameworks and brings general improvements at a negligible additional computation cost. Extensive experiments on CIFAR-100, ImageNet-2012, and MS-COCO demonstrate the effectiveness of our method. Our code is available at https://github.com/zhengli97/CTKD.

cs.CV↗

arXiv2022

ReCode: Robustness Evaluation of Code Generation Models

Code generation models have achieved impressive performance. However, they tend to be brittle as slight edits to a prompt could lead to very different generations; these robustness properties, critical for user experience when deployed in real-life applications, are not well understood. Most existing works on robustness in text or code tasks have focused on classification, while robustness in generation tasks is an uncharted area and to date there is no comprehensive benchmark for robustness in code generation. In this paper, we propose ReCode, a comprehensive robustness evaluation benchmark for code generation models. We customize over 30 transformations specifically for code on docstrings, function and variable names, code syntax, and code format. They are carefully designed to be natural in real-life coding practice, preserve the original semantic meaning, and thus provide multifaceted assessments of a model's robustness performance. With human annotators, we verified that over 90% of the perturbed prompts do not alter the semantic meaning of the original prompt. In addition, we define robustness metrics for code generation models considering the worst-case behavior under each type of perturbation, taking advantage of the fact that executing the generated code can serve as objective evaluation. We demonstrate ReCode on SOTA models using HumanEval, MBPP, as well as function completion tasks derived from them. Interesting observations include: better robustness for CodeGen over InCoder and GPT-J; models are most sensitive to syntax perturbations; more challenging robustness evaluation on MBPP over HumanEval.

cs.LG↗

arXiv2022

Deep Multi-Modal Structural Equations For Causal Effect Estimation With Unstructured Proxies

Estimating the effect of intervention from observational data while accounting for confounding variables is a key task in causal inference. Oftentimes, the confounders are unobserved, but we have access to large amounts of additional unstructured data (images, text) that contain valuable proxy signal about the missing confounders. This paper argues that leveraging this unstructured data can greatly improve the accuracy of causal effect estimation. Specifically, we introduce deep multi-modal structural equations, a generative model for causal effect estimation in which confounders are latent variables and unstructured data are proxy variables. This model supports multiple multi-modal proxies (images, text) as well as missing data. We empirically demonstrate that our approach outperforms existing methods based on propensity scores and corrects for confounding using unstructured inputs on tasks in genomics and healthcare. Our methods can potentially support the use of large amounts of data that were previously not used in causal inference

cs.LG↗

arXiv2022

Testing for context-dependent changes in neural encoding in naturalistic experiments

We propose a decoding-based approach to detect context effects on neural codes in longitudinal neural recording data. The approach is agnostic to how information is encoded in neural activity, and can control for a variety of possible confounding factors present in the data. We demonstrate our approach by determining whether it is possible to decode location encoding from prefrontal cortex in the mouse and, further, testing whether the encoding changes due to task engagement.

stat.ML↗

arXiv2022

Disentangling Task Relations for Few-shot Text Classification via Self-Supervised Hierarchical Task Clustering

Few-Shot Text Classification (FSTC) imitates humans to learn a new text classifier efficiently with only few examples, by leveraging prior knowledge from historical tasks. However, most prior works assume that all the tasks are sampled from a single data source, which cannot adapt to real-world scenarios where tasks are heterogeneous and lie in different distributions. As such, existing methods may suffer from their globally knowledge-shared mechanisms to handle the task heterogeneity. On the other hand, inherent task relation are not explicitly captured, making task knowledge unorganized and hard to transfer to new tasks. Thus, we explore a new FSTC setting where tasks can come from a diverse range of data sources. To address the task heterogeneity, we propose a self-supervised hierarchical task clustering (SS-HTC) method. SS-HTC not only customizes cluster-specific knowledge by dynamically organizing heterogeneous tasks into different clusters in hierarchical levels but also disentangles underlying relations between tasks to improve the interpretability. Extensive experiments on five public FSTC benchmark datasets demonstrate the effectiveness of SS-HTC.

cs.CL↗

arXiv2022

Unified Loss of Pair Similarity Optimization for Vision-Language Retrieval

There are two popular loss functions used for vision-language retrieval, i.e., triplet loss and contrastive learning loss, both of them essentially minimize the difference between the similarities of negative pairs and positive pairs. More specifically, Triplet loss with Hard Negative mining (Triplet-HN), which is widely used in existing retrieval models to improve the discriminative ability, is easy to fall into local minima in training. On the other hand, Vision-Language Contrastive learning loss (VLC), which is widely used in the vision-language pre-training, has been shown to achieve significant performance gains on vision-language retrieval, but the performance of fine-tuning with VLC on small datasets is not satisfactory. This paper proposes a unified loss of pair similarity optimization for vision-language retrieval, providing a powerful tool for understanding existing loss functions. Our unified loss includes the hard sample mining strategy of VLC and introduces the margin used by the triplet loss for better similarity separation. It is shown that both Triplet-HN and VLC are special forms of our unified loss. Compared with the Triplet-HN, our unified loss has a fast convergence speed. Compared with the VLC, our unified loss is more discriminative and can provide better generalization in downstream fine-tuning tasks. Experiments on image-text and video-text retrieval benchmarks show that our unified loss can significantly improve the performance of the state-of-the-art retrieval models.

cs.CV↗

arXiv2022

Safety-centric and Smart Outdoor Workplace: A New Research Direction and Its Technical Challenges

Despite the fact that outside is becoming the frontier of indoor workplaces, a large amount of real-world work like road construction has to be done by outdoor human activities in open areas. Given the promise of the smart workplace in various aspects including productivity and safety, we decided to employ smart workplace technologies for a collaborative outdoor project both to improve the work efficiency and to reduce the worker injuries. Nevertheless, our trials on smart workplace implementation have encountered a few problems ranging from the theoretical confusion among different stakeholders, to the technical difficulties in extending underground devices' lifespan. This triggers our rethinking of and discussions about "smart workplace". Eventually, considering the unique characteristics of outdoor work (e.g., more sophisticated workflows and more safety-related situations than office work), we argue that "safety-centric and smart outdoor workplace" deserves dedicated research attentions and efforts under the umbrella discipline of smart environment. In addition, the identified technical challenges can in turn drive different research dimensions of such a distinguishing topic.

cs.HC↗

arXiv2022

Academic Search Engines: Constraints, Bugs, and Recommendation

Background: Academic search engines (i.e., digital libraries and indexers) play an increasingly important role in systematic reviews however these engines do not seem to effectively support such reviews, e.g., researchers confront usability issues with the engines when conducting their searches. Aims: To investigate whether the usability issues are bugs (i.e., faults in the search engines) or constraints, and to provide recommendations to search-engine providers and researchers on how to tackle these issues. Method: Using snowball-sampling from tertiary studies, we identify a set of 621 secondary studies in software engineering. By physically re-attempting the searches for all of these 621 studies, we effectively conduct regression testing for 42 search engines. Results: We identify 13 bugs for eight engines, and also identify other constraints. We provide recommendations for tackling these issues. Conclusions: There is still a considerable gap between the search-needs of researchers and the usability of academic search engines. It is not clear whether search-engine developers are aware of this gap. Also, the evaluation, by academics, of academic search engines has not kept pace with the development, by search-engine providers, of those search engines. Thus, the gap between evaluation and development makes it harder to properly understand the gap between the search-needs of researchers and search-features of the search engines.

cs.SE↗

arXiv2022

On Kubernetes-aided Federated Database Systems

Cloud computing has made federated database systems (FDBS) significantly more practical to implement than in the past. As part of a recent Web-based Geographic Information System (WebGIS) project, we are employing cloud-native technologies (from the container ecosystem) to develop a federated database (DB) infrastructure, to help manage and utilise the distributed and various geospatial data. Unfortunately, there seem to be inherent challenges and complexity of applying the container and Kubernetes technologies to building and running DB systems. Considering that most of the geospatial and theme data are pre-obtained and fixed in our WebGIS project, we decided to focus on the read-only user queries and still resort to Kubernetes to implement an FDBS instance to use. Unlike the de facto practices (e.g., using the StatefulSets mechanism, extending Kuberentes APIs, or employing KubeFed), our solution for Kubernetes-aided FDBS simplifies the tech stack by investigating the fractal object of federated data management, inclusively containerising DB instances, and using the lightweight Deployment mechanism to handle stateless DB containers. Overall, this research not only reveals an easy-to-implement approach to constructing read-only components in a fully-fledged FDBS, but also proposes and demonstrates a novel methodology for FDBS investigations.

cs.DC↗

arXiv2022

Image-Text Retrieval with Binary and Continuous Label Supervision

Most image-text retrieval work adopts binary labels indicating whether a pair of image and text matches or not. Such a binary indicator covers only a limited subset of image-text semantic relations, which is insufficient to represent relevance degrees between images and texts described by continuous labels such as image captions. The visual-semantic embedding space obtained by learning binary labels is incoherent and cannot fully characterize the relevance degrees. In addition to the use of binary labels, this paper further incorporates continuous pseudo labels (generally approximated by text similarity between captions) to indicate the relevance degrees. To learn a coherent embedding space, we propose an image-text retrieval framework with Binary and Continuous Label Supervision (BCLS), where binary labels are used to guide the retrieval model to learn limited binary correlations, and continuous labels are complementary to the learning of image-text semantic relations. For the learning of binary labels, we improve the common Triplet ranking loss with Soft Negative mining (Triplet-SN) to improve convergence. For the learning of continuous labels, we design Kendall ranking loss inspired by Kendall rank correlation coefficient (Kendall), which improves the correlation between the similarity scores predicted by the retrieval model and the continuous labels. To mitigate the noise introduced by the continuous pseudo labels, we further design Sliding Window sampling and Hard Sample mining strategy (SW-HS) to alleviate the impact of noise and reduce the complexity of our framework to the same order of magnitude as the triplet ranking loss. Extensive experiments on two image-text retrieval benchmarks demonstrate that our method can improve the performance of state-of-the-art image-text retrieval models.

cs.CV↗

arXiv2022

Learning to Sample and Aggregate: Few-shot Reasoning over Temporal Knowledge Graphs

In this paper, we investigate a realistic but underexplored problem, called few-shot temporal knowledge graph reasoning, that aims to predict future facts for newly emerging entities based on extremely limited observations in evolving graphs. It offers practical value in applications that need to derive instant new knowledge about new entities in temporal knowledge graphs (TKGs) with minimal supervision. The challenges mainly come from the few-shot and time shift properties of new entities. First, the limited observations associated with them are insufficient for training a model from scratch. Second, the potentially dynamic distributions from the initially observable facts to the future facts ask for explicitly modeling the evolving characteristics of new entities. We correspondingly propose a novel Meta Temporal Knowledge Graph Reasoning (MetaTKGR) framework. Unlike prior work that relies on rigid neighborhood aggregation schemes to enhance low-data entity representation, MetaTKGR dynamically adjusts the strategies of sampling and aggregating neighbors from recent facts for new entities, through temporally supervised signals on future facts as instant feedback. Besides, such a meta temporal reasoning procedure goes beyond existing meta-learning paradigms on static knowledge graphs that fail to handle temporal adaptation with large entity variance. We further provide a theoretical analysis and propose a temporal adaptation regularizer to stabilize the meta temporal reasoning over time. Empirically, extensive experiments on three real-world TKGs demonstrate the superiority of MetaTKGR over state-of-the-art baselines by a large margin.

cs.LG↗

arXiv2022

Short Text Pre-training with Extended Token Classification for E-commerce Query Understanding

E-commerce query understanding is the process of inferring the shopping intent of customers by extracting semantic meaning from their search queries. The recent progress of pre-trained masked language models (MLM) in natural language processing is extremely attractive for developing effective query understanding models. Specifically, MLM learns contextual text embedding via recovering the masked tokens in the sentences. Such a pre-training process relies on the sufficient contextual information. It is, however, less effective for search queries, which are usually short text. When applying masking to short search queries, most contextual information is lost and the intent of the search queries may be changed. To mitigate the above issues for MLM pre-training on search queries, we propose a novel pre-training task specifically designed for short text, called Extended Token Classification (ETC). Instead of masking the input text, our approach extends the input by inserting tokens via a generator network, and trains a discriminator to identify which tokens are inserted in the extended input. We conduct experiments in an E-commerce store to demonstrate the effectiveness of ETC.

cs.CL↗