Search

Fonts are integral to creative endeavors, design processes, and artistic productions. The appropriate selection of a font can significantly enhance artwork and endow advertisements with a higher level of expressivity. Despite the availability of numerous diverse font designs online, traditional retrieval-based methods for font selection are increasingly being supplanted by generation-based approaches. These newer methods offer enhanced flexibility, catering to specific user preferences and capturing unique stylistic impressions. However, current impression font techniques based on Generative Adversarial Networks (GANs) necessitate the utilization of multiple auxiliary losses to provide guidance during generation. Furthermore, these methods commonly employ weighted summation for the fusion of impression-related keywords. This leads to generic vectors with the addition of more impression keywords, ultimately lacking in detail generation capacity. In this paper, we introduce a diffusion-based method, termed \ourmethod, to generate fonts that vividly embody specific impressions, utilizing an input consisting of a single letter and a set of descriptive impression keywords. The core innovation of \ourmethod lies in the development of dual cross-attention modules, which process the characteristics of the letters and impression keywords independently but synergistically, ensuring effective integration of both types of information. Our experimental results, conducted on the MyFonts dataset, affirm that this method is capable of producing realistic, vibrant, and high-fidelity fonts that are closely aligned with user specifications. This confirms the potential of our approach to revolutionize font generation by accommodating a broad spectrum of user-driven design requirements. Our code is publicly available at \url{https://github.com/leitro/GRIF-DM}., Comment: Accepted to ECAI2024

We propose a new graph-based framework to reveal relationships among motivations, emotions and actions explicitly given natural language texts. A directed acyclic graph is designed to describe human's nature. Nurture beliefs are incorporated to connect outside events and the human's nature graph. No annotation resources are required due to the power of large language models. Amazon Fine Foods Reviews dataset is used as corpus and food-related motivations are focused. Totally 92,990 relationship graphs are generated, of which 63% make logical sense. We make further analysis to investigate error types for optimization direction in future research.

Integrated super-resolution sensing and communication (ISSAC) has emerged as a promising technology to achieve extremely high precision sensing for those key parameters, such as the angles of the sensing targets. In this paper, we propose an efficient channel estimation scheme enabled by ISSAC for millimeter wave (mmWave) and TeraHertz (THz) systems with a hybrid analog/digital beamforming architecture, where both the pilot overhead and the cost of radio frequency (RF) chains are significantly reduced. The key idea is to exploit the fact that subspace-based super-resolution algorithms such as multiple signal classification (MUSIC) can estimate channel parameters accurately without requiring dedicate a priori known pilots. In particular, the proposed method consists of two stages. First, the angles of the multi-path channel components are estimated in a pilot-free manner during the transmission of data symbols. Second, the multi-path channel coefficients are estimated with very few pilots. Compared to conventional channel estimation schemes that rely solely on channel training, our approach requires the estimation of much fewer parameters in the second stage. Furthermore, with channel multi-path angles obtained, the beamforming gain can be achieved when pilots are sent to estimate the channel path gains. To comprehensively investigate the performance of the proposed scheme, we consider both the basic line-of-sight (LoS) channels and more general multi-path channels. We compare the performance of the minimum mean square error (MMSE) of channel estimation and the resulting beamforming gains of our proposed scheme with the traditional scheme that rely exclusively on channel training. It is demonstrated that our proposed method significantly outperforms the benchmarking scheme. Simulation results are presented to validate our theoretical findings., Comment: 13 pages, 8 figures

Transformer models have emerged as potent solutions to a wide array of multidisciplinary challenges. The deployment of Transformer architectures is significantly hindered by their extensive computational and memory requirements, necessitating the reliance on advanced efficient distributed training methodologies. Prior research has delved into the performance bottlenecks associated with distributed training, aiming to unravel these bottlenecks and suggest optimization directions. However, such analyses often overlook three aspects unique to Transformer models: the specialized architecture, the dependency on various distributed strategies, and the requirement to balance computational and memory overhead. This paper aims to bridge this gap by offering a comprehensive examination of the performance bottlenecks inherent in distributed training of Transformer models, leveraging both theoretical analysis and empirical investigation. We propose an analytical framework tailored to these unique aspects of Transformers, facilitating a holistic evaluation of model architectures, distributed strategies, and resource consumption. Based on this analytical framework, we conduct a comparative analysis of theoretical performances and further systematically explore how various distributed training strategies fare in real-world scenarios. Most of the experimental results can be well explained by the analytical outcomes derived from the analytical framework. Notably, our findings suggest an advantage of pipeline parallelism over data parallelism for Transformer models. Moreover, we shed light on some unexpected outcomes, such as the potential for increased total memory overhead due to suboptimal model partitioning within pipeline parallelism. Additionally, we underscore the significance of communication block size and waiting time to further enhance performance.

Vision Transformers (ViTs) have exhibited exceptional performance across diverse computer vision tasks, while their substantial parameter size incurs significantly increased memory and computational demands, impeding effective inference on resource-constrained devices. Quantization has emerged as a promising solution to mitigate these challenges, yet existing methods still suffer from significant accuracy loss at low-bit. We attribute this issue to the distinctive distributions of post-LayerNorm and post-GELU activations within ViTs, rendering conventional hardware-friendly quantizers ineffective, particularly in low-bit scenarios. To address this issue, we propose a novel framework called Activation-Distribution-Friendly post-training Quantization for Vision Transformers, ADFQ-ViT. Concretely, we introduce the Per-Patch Outlier-aware Quantizer to tackle irregular outliers in post-LayerNorm activations. This quantizer refines the granularity of the uniform quantizer to a per-patch level while retaining a minimal subset of values exceeding a threshold at full-precision. To handle the non-uniform distributions of post-GELU activations between positive and negative regions, we design the Shift-Log2 Quantizer, which shifts all elements to the positive region and then applies log2 quantization. Moreover, we present the Attention-score enhanced Module-wise Optimization which adjusts the parameters of each quantizer by reconstructing errors to further mitigate quantization error. Extensive experiments demonstrate ADFQ-ViT provides significant improvements over various baselines in image classification, object detection, and instance segmentation tasks at 4-bit. Specifically, when quantizing the ViT-B model to 4-bit, we achieve a 10.23% improvement in Top-1 accuracy on the ImageNet dataset., Comment: 28 pages,9 figures

Large-scale Text-to-Image (T2I) diffusion models demonstrate significant generation capabilities based on textual prompts. Based on the T2I diffusion models, text-guided image editing research aims to empower users to manipulate generated images by altering the text prompts. However, existing image editing techniques are prone to editing over unintentional regions that are beyond the intended target area, primarily due to inaccuracies in cross-attention maps. To address this problem, we propose Localization-aware Inversion (LocInv), which exploits segmentation maps or bounding boxes as extra localization priors to refine the cross-attention maps in the denoising phases of the diffusion process. Through the dynamic updating of tokens corresponding to noun words in the textual input, we are compelling the cross-attention maps to closely align with the correct noun and adjective words in the text prompt. Based on this technique, we achieve fine-grained image editing over particular objects while preventing undesired changes to other regions. Our method LocInv, based on the publicly available Stable Diffusion, is extensively evaluated on a subset of the COCO dataset, and consistently obtains superior results both quantitatively and qualitatively.The code will be released at https://github.com/wangkai930418/DPL, Comment: Accepted by CVPR 2024 Workshop AI4CC

Document understanding models have recently demonstrated remarkable performance by leveraging extensive collections of user documents. However, since documents often contain large amounts of personal data, their usage can pose a threat to user privacy and weaken the bonds of trust between humans and AI services. In response to these concerns, legislation advocating ``the right to be forgotten" has recently been proposed, allowing users to request the removal of private information from computer systems and neural network models. A novel approach, known as machine unlearning, has emerged to make AI models forget about a particular class of data. In our research, we explore machine unlearning for document classification problems, representing, to the best of our knowledge, the first investigation into this area. Specifically, we consider a realistic scenario where a remote server houses a well-trained model and possesses only a small portion of training data. This setup is designed for efficient forgetting manipulation. This work represents a pioneering step towards the development of machine unlearning methods aimed at addressing privacy concerns in document analysis applications. Our code is publicly available at \url{https://github.com/leitro/MachineUnlearning-DocClassification}., Comment: Accepted to ICDAR2024

The recent emergence of the hybrid quantum-classical neural network (HQCNN) architecture has garnered considerable attention due to the potential advantages associated with integrating quantum principles to enhance various facets of machine learning algorithms and computations. However, the current investigated serial structure of HQCNN, wherein information sequentially passes from one network to another, often imposes limitations on the trainability and expressivity of the network. In this study, we introduce a novel architecture termed Parallel Proportional Fusion of Quantum and Spiking Neural Networks (PPF-QSNN). The dataset information is simultaneously fed into both the spiking neural network and the variational quantum circuits, with the outputs amalgamated in proportion to their individual contributions. We systematically assess the impact of diverse PPF-QSNN parameters on network performance for image classification, aiming to identify the optimal configuration. Numerical results on the MNIST dataset unequivocally illustrate that our proposed PPF-QSNN outperforms both the existing spiking neural network and the serial quantum neural network across metrics such as accuracy, loss, and robustness. This study introduces a novel and effective amalgamation approach for HQCNN, thereby laying the groundwork for the advancement and application of quantum advantage in artificial intelligent computations.

Given unlabelled datasets containing both old and new categories, generalized category discovery (GCD) aims to accurately discover new classes while correctly classifying old classes, leveraging the class concepts learned from labeled samples. Current GCD methods only use a single visual modality of information, resulting in poor classification of visually similar classes. As a different modality, text information can provide complementary discriminative information, which motivates us to introduce it into the GCD task. However, the lack of class names for unlabelled data makes it impractical to utilize text information. To tackle this challenging problem, in this paper, we propose a Text Embedding Synthesizer (TES) to generate pseudo text embeddings for unlabelled samples. Specifically, our TES leverages the property that CLIP can generate aligned vision-language features, converting visual embeddings into tokens of the CLIP's text encoder to generate pseudo text embeddings. Besides, we employ a dual-branch framework, through the joint learning and instance consistency of different modality branches, visual and semantic information mutually enhance each other, promoting the interaction and fusion of visual and text knowledge. Our method unlocks the multi-modal potentials of CLIP and outperforms the baseline methods by a large margin on all GCD benchmarks, achieving new state-of-the-art. The code will be released at https://github.com/enguangW/GET .

Driven by the demand for energy-efficient employment of deep neural networks, early-exit methods have experienced a notable increase in research attention. These strategies allow for swift predictions by making decisions early in the network, thereby conserving computation time and resources. However, so far the early-exit networks have only been developed for stationary data distributions, which restricts their application in real-world scenarios with continuous non-stationary data. This study aims to explore the continual learning of the early-exit networks. We adapt existing continual learning methods to fit with early-exit architectures and investigate their behavior in the continual setting. We notice that early network layers exhibit reduced forgetting and can outperform standard networks even when using significantly fewer resources. Furthermore, we analyze the impact of task-recency bias on early-exit inference and propose Task-wise Logits Correction (TLC), a simple method that equalizes this bias and improves the network performance for every given compute budget in the class-incremental setting. We assess the accuracy and computational cost of various continual learning techniques enhanced with early-exits and TLC across standard class-incremental learning benchmarks such as 10 split CIFAR100 and ImageNetSubset and show that TLC can achieve the accuracy of the standard methods using less than 70\% of their computations. Moreover, at full computational budget, our method outperforms the accuracy of the standard counterparts by up to 15 percentage points. Our research underscores the inherent synergy between early-exit networks and continual learning, emphasizing their practical utility in resource-constrained environments.

Large language models (LLMs) have demonstrated state-of-the-art performance across various tasks. However, the latency of inference and the large GPU memory consumption of LLMs restrict their deployment performance. Recently, there have been some efficient attempts to quantize LLMs, yet inference with large batch size or long sequence still has the issue of being compute-bound. Fine-grained quantization methods have showcased their proficiency in achieving low-bit quantization for LLMs, while requiring FP16 data type for linear layer computations, which is time-consuming when dealing with large batch size or long sequence. In this paper, we introduce a method called FlattenQuant, which significantly reduces the maximum value of the tensor by flattening the large channels in the tensor, to achieve low bit per-tensor quantization with minimal accuracy loss. Our experiments show that FlattenQuant can directly use 4 bits to achieve 48.29% of the linear layer calculation in LLMs, with the remaining layers using 8 bits. The 4-bit matrix multiplication introduced in the FlattenQuant method can effectively address the compute-bound caused by large matrix calculation. Our work achieves up to 2$\times$ speedup and 2.3$\times$ memory reduction for LLMs with negligible loss in accuracy.

Let $\theta$ be an irrational number of bounded type. We prove that all capture components in the parameter space of cubic polynomials $f_a(z)=e^{2\pi i\theta}z+a z^2+z^3$, where $a\in\mathbb{C}$, are Jordan domains., Comment: 15 pages, 2 figures

Large language models (LLMs) such as GPT-3, OPT, and LLaMA have demonstrated remarkable accuracy in a wide range of tasks. However, training these models can incur significant expenses, often requiring tens of thousands of GPUs for months of continuous operation. Typically, this training is carried out in specialized GPU clusters equipped with homogeneous high-speed Remote Direct Memory Access (RDMA) network interface cards (NICs). The acquisition and maintenance of such dedicated clusters is challenging. Current LLM training frameworks, like Megatron-LM and Megatron-DeepSpeed, focus primarily on optimizing training within homogeneous cluster settings. In this paper, we introduce Holmes, a training framework for LLMs that employs thoughtfully crafted data and model parallelism strategies over the heterogeneous NIC environment. Our primary technical contribution lies in a novel scheduling method that intelligently allocates distinct computational tasklets in LLM training to specific groups of GPU devices based on the characteristics of their connected NICs. Furthermore, our proposed framework, utilizing pipeline parallel techniques, demonstrates scalability to multiple GPU clusters, even in scenarios without high-speed interconnects between nodes in distinct clusters. We conducted comprehensive experiments that involved various scenarios in the heterogeneous NIC environment. In most cases, our framework achieves performance levels close to those achievable with homogeneous RDMA-capable networks (InfiniBand or RoCE), significantly exceeding training efficiency within the pure Ethernet environment. Additionally, we verified that our framework outperforms other mainstream LLM frameworks under heterogeneous NIC environment in terms of training efficiency and can be seamlessly integrated with them., Comment: 12 pages

The Quantum Approximate Optimization Algorithm (QAOA), a pivotal paradigm in the realm of variational quantum algorithms (VQAs), offers promising computational advantages for tackling combinatorial optimization problems. Well-defined initial circuit parameters, responsible for preparing a parameterized quantum state encoding the solution, play a key role in optimizing QAOA. However, classical optimization techniques encounter challenges in discerning optimal parameters that align with the optimal solution. In this work, we propose a hybrid optimization approach that integrates Gated Recurrent Units (GRU), Convolutional Neural Networks (CNN), and a bilinear strategy as an innovative alternative to conventional optimizers for predicting optimal parameters of QAOA circuits. GRU serves to stochastically initialize favorable parameters for depth-1 circuits, while CNN predicts initial parameters for depth-2 circuits based on the optimized parameters of depth-1 circuits. To assess the efficacy of our approach, we conducted a comparative analysis with traditional initialization methods using QAOA on Erd\H{o}s-R\'enyi graph instances, revealing superior optimal approximation ratios. We employ the bilinear strategy to initialize QAOA circuit parameters at greater depths, with reference parameters obtained from GRU-CNN optimization. This approach allows us to forecast parameters for a depth-12 QAOA circuit, yielding a remarkable approximation ratio of 0.998 across 10 qubits, which surpasses that of the random initialization strategy and the PPN2 method at a depth of 10. The proposed hybrid GRU-CNN bilinear optimization method significantly improves the effectiveness and accuracy of parameters initialization, offering a promising iterative framework for QAOA that elevates its performance.

Novel topological properties that arose in the periodically driven system are unique, in which there are two kinds of quasienergy gaps, the zero quasienergy gap and the $\pi$ quasienergy gap. The corresponding edge modes would traverse either the zero quasienergy gap or the $\pi$ quasienergy gap, or traverse both two quasienergy gaps. And the characterization of these two kinds of edge modes might not be the same. However, in this paper, we find that both the zero edge modes and the $\pi$ edge modes in the Floquet Chern insulators can be characterized by the same topological invariant, where the corresponding Dirac mass term is periodically kicked. Particularly, we take the periodically kicked Qi-Wu-Zhang model as an illustrative example. In this model, the topology is characterized by the Floquet Chern number $C_F$, and there are six different topological phases in total, denoted as $C_F=\{-1_0,-2,-1_\pi,1_\pi,2,1_0\}$. Furthermore, we find that the Floquet operator associated with the periodically kicked Qi-Wu-Zhang model reduces to a Dirac Hamiltonian in the low-energy limit. Then, the phase diagram is uncovered by examining the topology of this Dirac Hamiltonian. Additionally, we explore the orders of topological phase transitions in the context of Floquet stationary states by analyzing the von Neumann entropy of these states. Our work provides further insights into the topological phases in periodically driven systems., Comment: 10 pages, 7 figures

Recent advancements in unsupervised protein language models (ProteinLMs), like ESM-1b and ESM-2, have shown promise in different protein prediction tasks. However, these models face challenges due to their high computational demands, significant memory needs, and latency, restricting their usage on devices with limited resources. To tackle this, we explore post-training quantization (PTQ) for ProteinLMs, focusing on ESMFold, a simplified version of AlphaFold based on ESM-2 ProteinLM. Our study is the first attempt to quantize all weights and activations of ProteinLMs. We observed that the typical uniform quantization method performs poorly on ESMFold, causing a significant drop in TM-Score when using 8-bit quantization. We conducted extensive quantization experiments, uncovering unique challenges associated with ESMFold, particularly highly asymmetric activation ranges before Layer Normalization, making representation difficult using low-bit fixed-point formats. To address these challenges, we propose a new PTQ method for ProteinLMs, utilizing piecewise linear quantization for asymmetric activation values to ensure accurate approximation. We demonstrated the effectiveness of our method in protein structure prediction tasks, demonstrating that ESMFold can be accurately quantized to low-bit widths without compromising accuracy. Additionally, we applied our method to the contact prediction task, showcasing its versatility. In summary, our study introduces an innovative PTQ method for ProteinLMs, addressing specific quantization challenges and potentially leading to the development of more efficient ProteinLMs with significant implications for various protein-related applications., Comment: 8 pages, 4 figures

Large-scale text-to-image generative models have been a ground-breaking development in generative AI, with diffusion models showing their astounding ability to synthesize convincing images following an input text prompt. The goal of image editing research is to give users control over the generated images by modifying the text prompt. Current image editing techniques are susceptible to unintended modifications of regions outside the targeted area, such as on the background or on distractor objects which have some semantic or visual relationship with the targeted object. According to our experimental findings, inaccurate cross-attention maps are at the root of this problem. Based on this observation, we propose Dynamic Prompt Learning (DPL) to force cross-attention maps to focus on correct noun words in the text prompt. By updating the dynamic tokens for nouns in the textual input with the proposed leakage repairment losses, we achieve fine-grained image editing over particular objects while preventing undesired changes to other image regions. Our method DPL, based on the publicly available Stable Diffusion, is extensively evaluated on a wide range of images, and consistently obtains superior results both quantitatively (CLIP score, Structure-Dist) and qualitatively (on user-evaluation). We show improved prompt editing results for Word-Swap, Prompt Refinement, and Attention Re-weighting, especially for complex multi-object scenes., Comment: Neurips 2023. The code page: https://github.com/wangkai930418/DPL

For post-critically finite rational maps, we give a sufficient condition for the existence of disjoint type renormalizations by using multicurves. We also give examples satisfying the condition., Comment: 22 pages, 7 figures

The principle underlying most existing continual learning (CL) methods is to prioritize stability by penalizing changes in parameters crucial to old tasks, while allowing for plasticity in other parameters. The importance of weights for each task can be determined either explicitly through learning a task-specific mask during training (e.g., parameter isolation-based approaches) or implicitly by introducing a regularization term (e.g., regularization-based approaches). However, all these methods assume that the importance of weights for each task is unknown prior to data exposure. In this paper, we propose ScrollNet as a scrolling neural network for continual learning. ScrollNet can be seen as a dynamic network that assigns the ranking of weight importance for each task before data exposure, thus achieving a more favorable stability-plasticity tradeoff during sequential task learning by reassigning this ranking for different tasks. Additionally, we demonstrate that ScrollNet can be combined with various CL methods, including regularization-based and replay-based approaches. Experimental results on CIFAR100 and TinyImagenet datasets show the effectiveness of our proposed method. We release our code at https://github.com/FireFYF/ScrollNet.git., Comment: Accepted at Visual Continual Learning workshop (ICCV2023)

Recently, much attention has been paid to uncovering the influence of dissipation on a quantum system, particularly on how the non-Hermitian (NH) terms modify the band topology of topological materials and reshape the profile of the wavefunctions of a system (or the NH skin effect). In this paper, a specific NH skin effect that induced by local dissipation is studied for chiral edge modes, in which the NH term corresponds to the local imaginary Fermi velocity of the chiral edge modes. By solving the NH Schr\"{o}dinger equation of the non-Hermitian chiral edge modes (nhCEs) with complex Fermi velocity, we uncovered the remarkable complex spectra and the wavefunctions of the nhCEs. We find that the complex spectra of these modes is a straight line in the topological materials, and its chirality can separates the modes with positive energy from those with negative energy, in which they are localized at different positions. We also studied the nhCEs at the boundary of 2-dimensional (2D) topological materials, the 2D $p$-wave superconductor and the Qi-Wu-Zhang model, in which the general law of nhCEs was verified. We expect that our findings will pave the way for researching the transport properties of the chiral edge modes in the non-equilibrium context., Comment: 7 pages, 6 figures

In recent years, commonsense reasoning has received more and more attention from academic community. We propose a new lexical inference task, Mental and Physical Classification (MPC), to handle commonsense reasoning in a reasoning graph. Mental words relate to mental activities, which fall into six categories: Emotion, Need, Perceiving, Reasoning, Planning and Personality. Physical words describe physical attributes of an object, like color, hardness, speed and malleability. A BERT model is fine-tuned for this task and active learning algorithm is adopted in the training framework to reduce the required annotation resources. The model using ENTROPY strategy achieves satisfactory accuracy and requires only about 300 labeled words. We also compare our result with SentiWordNet to check the difference between MPC and subjectivity classification task in sentiment analysis., Comment: 10 pages, 2 figures

Nowadays, smart contracts have become increasingly popular and, as with software development in general, testing is the standard method for verifying their correctness. However, smart contracts require a higher level of certainty regarding correctness because they are diffcult to modify once deployed and errors can result in significant financial losses. Therefore, formal verification is essential. In this article, we present our search for a tool capable of formally verifying a real-world smart contract written in a recent version of Solidity.

Let $f_\theta(z)=e^{2\pi i\theta}z+z^2$ be the quadratic polynomial having an indifferent fixed point at the origin. For any bounded type irrational number $\theta\in\mathbb{R}\setminus\mathbb{Q}$ and any rational number $\nu\in\mathbb{Q}$, we prove that $f_\theta$ and $f_\nu$ are conformally mateable, and that the mating is unique up to conjugacy by a M\"{o}bius map. This gives an affirmative (partial) answer to a question raised by Milnor in 2004. A crucial ingredient in the proof relies on an expansive property when iterating certain rational maps near Siegel disk boundaries. Combining this with the expanding property in repelling petals of parabolic points, we also prove that the Julia sets of a class of Siegel rational maps with parabolic points are locally connected., Comment: 40 pages, 9 figures

Recommendation models that utilize unique identities (IDs) to represent distinct users and items have been state-of-the-art (SOTA) and dominated the recommender systems (RS) literature for over a decade. Meanwhile, the pre-trained modality encoders, such as BERT and ViT, have become increasingly powerful in modeling the raw modality features of an item, such as text and images. Given this, a natural question arises: can a purely modality-based recommendation model (MoRec) outperforms or matches a pure ID-based model (IDRec) by replacing the itemID embedding with a SOTA modality encoder? In fact, this question was answered ten years ago when IDRec beats MoRec by a strong margin in both recommendation accuracy and efficiency. We aim to revisit this `old' question and systematically study MoRec from several aspects. Specifically, we study several sub-questions: (i) which recommendation paradigm, MoRec or IDRec, performs better in practical scenarios, especially in the general setting and warm item scenarios where IDRec has a strong advantage? does this hold for items with different modality features? (ii) can the latest technical advances from other communities (i.e., natural language processing and computer vision) translate into accuracy improvement for MoRec? (iii) how to effectively utilize item modality representation, can we use it directly or do we have to adjust it with new data? (iv) are there some key challenges for MoRec to be solved in practical applications? To answer them, we conduct rigorous experiments for item recommendations with two popular modalities, i.e., text and vision. We provide the first empirical evidence that MoRec is already comparable to its IDRec counterpart with an expensive end-to-end training method, even for warm item recommendation. Our results potentially imply that the dominance of IDRec in the RS field may be greatly challenged in the future.

We demonstrate the stable and flexible light delivery of multi-{\mu}J, sub-200-fs pulses over a ~10-m-long vacuumized anti-resonant hollow-core fiber (AR-HCF), which was successfully used for high-performance pulse synchronization. Compared with the pulse train launched into the AR-HCF, the transmitted pulse train out of the fiber exhibits excellent stabilities in pulse power and spectrum, with pointing stability largely improved. The walk-off between the fiber-delivery and the other free-space-propagation pulse trains, in an open loop, was measured to be <6 fs root-mean-square (RMS) over 90 minutes, corresponding to a relative optical-path variation of <2x10-7. This walk-off can be further suppressed to ~2 fs RMS simply using an active control loop, highlighting the great application potentials of this AR-HCF set-up in large-scale laser and accelerator facilities.

We classify the dissipative topological insulators (TIs) with edge dark states (EDS) by using the 38-fold way of non-Hermitian systems in this paper. The dissipative dynamics of these quadratic open fermionic systems is captured by a non-Hermitian single-particle matrix which contains both the internal dynamics and the dissipation, refereed to as damping matrix $X$. And the dark states in these systems are the eigenmodes of $X$ which the eigenvalues' imaginary part vanishes. However, there is a constraint on $X$, namely that the modes in which the eigenvalues' imaginary parts are positive are forbidden. In other words, the imaginary line-gap of $X$ is ill-defined, so the topological band theory classifying the dark states can not be applied to $X$. To reveal the topological protection of EDS, we propose the double damping matrix $\tilde{X} = \text{diag}\left( X, X^* \right)$, where the imaginary line-gap is well defined. Thus, the 38-fold way can be applied to $\tilde{X}$, and the topological protection of the EDS is uncovered. Different from previous studies of EDS in purely dissipative dynamics, the EDS in the dissipative TIs are robust against the inclusion of Hamiltonians. Furthermore, the topological classification of $\tilde{X}$ not only reflects the topological protection of EDS in dissipative TIs but also provides a paradigm to predict the appearance of EDS in other open free fermionic systems., Comment: 8 pages, 4 figures

We propose a new method for exemplar-free class incremental training of ViTs. The main challenge of exemplar-free continual learning is maintaining plasticity of the learner without causing catastrophic forgetting of previously learned tasks. This is often achieved via exemplar replay which can help recalibrate previous task classifiers to the feature drift which occurs when learning new tasks. Exemplar replay, however, comes at the cost of retaining samples from previous tasks which for many applications may not be possible. To address the problem of continual ViT training, we first propose gated class-attention to minimize the drift in the final ViT transformer block. This mask-based gating is applied to class-attention mechanism of the last transformer block and strongly regulates the weights crucial for previous tasks. Importantly, gated class-attention does not require the task-ID during inference, which distinguishes it from other parameter isolation methods. Secondly, we propose a new method of feature drift compensation that accommodates feature drift in the backbone when learning new tasks. The combination of gated class-attention and cascaded feature drift compensation allows for plasticity towards new tasks while limiting forgetting of previous ones. Extensive experiments performed on CIFAR-100, Tiny-ImageNet and ImageNet100 demonstrate that our exemplar-free method obtains competitive results when compared to rehearsal based ViT methods.

Self-supervised learning has been widely applied to train high-quality vision transformers. Unleashing their excellent performance on memory and compute constraint devices is therefore an important research topic. However, how to distill knowledge from one self-supervised ViT to another has not yet been explored. Moreover, the existing self-supervised knowledge distillation (SSKD) methods focus on ConvNet based architectures are suboptimal for ViT knowledge distillation. In this paper, we study knowledge distillation of self-supervised vision transformers (ViT-SSKD). We show that directly distilling information from the crucial attention mechanism from teacher to student can significantly narrow the performance gap between both. In experiments on ImageNet-Subset and ImageNet-1K, we show that our method AttnDistill outperforms existing self-supervised knowledge distillation (SSKD) methods and achieves state-of-the-art k-NN accuracy compared with self-supervised learning (SSL) methods learning from scratch (with the ViT-S model). We are also the first to apply the tiny ViT-T model on self-supervised learning. Moreover, AttnDistill is independent of self-supervised learning algorithms, it can be adapted to ViT based SSL methods to improve the performance in future research. The code is here: https://github.com/wangkai930418/attndistill, Comment: BMVC 2022

We prove the existence of rational maps having smooth degenerate Herman rings. This answers a question of Eremenko affirmatively. The proof is based on the construction of smooth Siegel disks by Avila, Buff and Ch\'{e}ritat as well as the classical Siegel-to-Herman quasiconformal surgery. A crucial ingredient in the proof is the surgery's continuity, which relies on the control of the loss of the area of quadratic filled-in Julia sets by Buff and Ch\'{e}ritat. As a by-product, we prove the existence of rational maps having a nowhere dense Julia set of positive area for which these maps have no irrationally indifferent periodic points, no Herman rings, and are not renormalizable., Comment: 32 pages, 3 figures

Accurate semantic segmentation models typically require significant computational resources, inhibiting their use in practical applications. Recent works rely on well-crafted lightweight models to achieve fast inference. However, these models cannot flexibly adapt to varying accuracy and efficiency requirements. In this paper, we propose a simple but effective slimmable semantic segmentation (SlimSeg) method, which can be executed at different capacities during inference depending on the desired accuracy-efficiency tradeoff. More specifically, we employ parametrized channel slimming by stepwise downward knowledge distillation during training. Motivated by the observation that the differences between segmentation results of each submodel are mainly near the semantic borders, we introduce an additional boundary guided semantic segmentation loss to further improve the performance of each submodel. We show that our proposed SlimSeg with various mainstream networks can produce flexible models that provide dynamic adjustment of computational cost and better performance than independent models. Extensive experiments on semantic segmentation benchmarks, Cityscapes and CamVid, demonstrate the generalization ability of our framework., Comment: 10 pages, 4 figures, accepted by ACMMM 2022, corrected the legend in fig.3