Search

Diffusion probabilistic models can generate high-quality samples. Yet, their sampling process requires numerous denoising steps, making it slow and computationally intensive. We propose to reduce the sampling cost by pruning a pretrained diffusion model into a mixture of efficient experts. First, we study the similarities between pairs of denoising timesteps, observing a natural clustering, even across different datasets. This suggests that rather than having a single model for all time steps, separate models can serve as ``experts'' for their respective time intervals. As such, we separately fine-tune the pretrained model on each interval, with elastic dimensions in depth and width, to obtain experts specialized in their corresponding denoising interval. To optimize the resource usage between experts, we introduce our Expert Routing Agent, which learns to select a set of proper network configurations. By doing so, our method can allocate the computing budget between the experts in an end-to-end manner without requiring manual heuristics. Finally, with a selected configuration, we fine-tune our pruned experts to obtain our mixture of efficient experts. We demonstrate the effectiveness of our method, DiffPruning, across several datasets, LSUN-Church, LSUN-Beds, FFHQ, and ImageNet, on the Latent Diffusion Model architecture., Comment: Accepted to the 18th European Conference on Computer Vision, ECCV 2024

We give a non-Archimedean characterization of K-semistability of log Fano cone singularities, and show that it agrees with the definition originally defined by Collins--Sz\'ekelyhidi. As an application, we show that to test K-semistability, it suffices to test special test configurations. We also show that special test configurations give rise to lc places of torus equivariant bounded complements., Comment: 28 pages

This paper introduces an approach to harness digital twin (DT) technology in the realm of integrated sensing and communications (ISAC) in the sixth-generation (6G) Internet-of-everything (IoE) applications. We consider moving targets in a vehicular network and use DT to track and predict the motion of the vehicles. After predicting the location of the vehicle at the next time slot, the DT designs the assignment and beamforming for each vehicle. The real time sensing information is then utilized to update and refine the DT, enabling further processing and decision-making. This model incorporates a dynamic Kalman gain, which is updated at each time slot based on the received echo signals. The state representation encompasses both vehicle motion information and the error matrix, with the posterior Cram\'er-Rao bound (PCRB) employed to assess sensing accuracy. We consider a network with two roadside units (RSUs), and the vehicles need to be allocated to one of them. To optimize the overall transmission rate while maintaining an acceptable sensing accuracy, an optimization problem is formulated. Since it is generally hard to solve the original problem, Lagrange multipliers and fractional programming are employed to simplify this optimization problem. To solve the simplified problem, this paper introduces both greedy and heuristic algorithms through optimizing both vehicle assignments and predictive beamforming. The optimized results are then transferred back to the real space for ISAC applications. Recognizing the computational complexity of the greedy and heuristic algorithms, a bidirectional long short-term memory (LSTM)-based recurrent neural network (RNN) is proposed for efficient beamforming design within the DT. Simulation results demonstrate the effectiveness of the DT-based ISAC network.

Recent breakthroughs in diffusion models have exhibited exceptional image-generation capabilities. However, studies show that some outputs are merely replications of training data. Such replications present potential legal challenges for model owners, especially when the generated content contains proprietary information. In this work, we introduce a straightforward yet effective method for detecting memorized prompts by inspecting the magnitude of text-conditional predictions. Our proposed method seamlessly integrates without disrupting sampling algorithms, and delivers high accuracy even at the first generation step, with a single generation per prompt. Building on our detection strategy, we unveil an explainable approach that shows the contribution of individual words or tokens to memorization. This offers an interactive medium for users to adjust their prompts. Moreover, we propose two strategies i.e., to mitigate memorization by leveraging the magnitude of text-conditional predictions, either through minimization during inference or filtering during training. These proposed strategies effectively counteract memorization while maintaining high-generation quality. Code is available at https://github.com/YuxinWenRick/diffusion_memorization., Comment: 16 pages, 9 figures, accepted as oral presentation in ICLR 2024

We propose the realization of interaction-driven insulators in periodically strained monolayer graphene and transition metal dichalcogenides (TMDs). By analyzing the tunable band structure and band geometry via strain, and performing extensive many-body exact diagonalization of a realistic model, we present compelling evidence for realizing various fractional Chern insulators in both strained monolayer graphene and TMDs. Our thorough analysis across different strain parameters, accounting for experimental variability, reveals that a broad spectrum of fractional Chern insulators, including the Laughlin states, Halperin 112, 332 and 111 states, and Chern number |C| = 2 states, can be stabilized in distinct regions of the phase diagram. These findings suggest that periodically strained monolayer graphene and TMDs provide promising platforms for engineering fractional Chern insulators., Comment: 10 pages, 5+7 figures. We corrected the model of PSTMDs from non-relativistic to massive Dirac electrons in a periodic magnetic field. Fig. 3 has been updated, and exact diagonalization recalculations confirm the FCI

In the era of 5G and beyond, the increasing complexity of wireless networks necessitates innovative frameworks for efficient management and deployment. Digital twins (DTs), embodying real-time monitoring, predictive configurations, and enhanced decision-making capabilities, stand out as a promising solution in this context. Within a time-series data-driven framework that effectively maps wireless networks into digital counterparts, encapsulated by integrated vertical and horizontal twinning phases, this study investigates the security challenges in distributed network DT systems, which potentially undermine the reliability of subsequent network applications such as wireless traffic forecasting. Specifically, we consider a minimal-knowledge scenario for all attackers, in that they do not have access to network data and other specialized knowledge, yet can interact with previous iterations of server-level models. In this context, we spotlight a novel fake traffic injection attack designed to compromise a distributed network DT system for wireless traffic prediction. In response, we then propose a defense mechanism, termed global-local inconsistency detection (GLID), to counteract various model poisoning threats. GLID strategically removes abnormal model parameters that deviate beyond a particular percentile range, thereby fortifying the security of network twinning process. Through extensive experiments on real-world wireless traffic datasets, our experimental evaluations show that both our attack and defense strategies significantly outperform existing baselines, highlighting the importance of security measures in the design and implementation of DTs for 5G and beyond network systems., Comment: Accepted by Internet of Things Journal (IoT-J). arXiv admin note: substantial text overlap with arXiv:2404.14389

We construct projective asymptotically good moduli spaces parametrizing boundary polarized CY surface pairs, which are projective slc Calabi-Yau pairs $(X,D)$ such that $D$ is ample and $X$ has dimension two. The moduli space provides a wall crossing between certain KSBA and K-moduli spaces and is the ample model of the Hodge line bundle. In the case of K3 surfaces with a non-symplectic automorphism, the moduli space gives a modular interpretation for the Baily--Borel compactification., Comment: 57 pages, comments are very welcome

Optimizing edge caching is crucial for the advancement of next-generation (nextG) wireless networks, ensuring high-speed and low-latency services for mobile users. Existing data-driven optimization approaches often lack awareness of the distribution of random data variables and focus solely on optimizing cache hit rates, neglecting potential reliability concerns, such as base station overload and unbalanced cache issues. This oversight can result in system crashes and degraded user experience. To bridge this gap, we introduce a novel digital twin-assisted optimization framework, called D-REC, which integrates reinforcement learning (RL) with diverse intervention modules to ensure reliable caching in nextG wireless networks. We first develop a joint vertical and horizontal twinning approach to efficiently create network digital twins, which are then employed by D-REC as RL optimizers and safeguards, providing ample datasets for training and predictive evaluation of our cache replacement policy. By incorporating reliability modules into a constrained Markov decision process, D-REC can adaptively adjust actions, rewards, and states to comply with advantageous constraints, minimizing the risk of network failures. Theoretical analysis demonstrates comparable convergence rates between D-REC and vanilla data-driven methods without compromising caching performance. Extensive experiments validate that D-REC outperforms conventional approaches in cache hit rate and load balancing while effectively enforcing predetermined reliability intervention modules., Comment: Accepted by IEEE Journal on Selected Areas in Communications (JSAC)

Federated learning (FL) enables multiple clients to collaboratively train machine learning models without revealing their private training data. In conventional FL, the system follows the server-assisted architecture (server-assisted FL), where the training process is coordinated by a central server. However, the server-assisted FL framework suffers from poor scalability due to a communication bottleneck at the server, and trust dependency issues. To address challenges, decentralized federated learning (DFL) architecture has been proposed to allow clients to train models collaboratively in a serverless and peer-to-peer manner. However, due to its fully decentralized nature, DFL is highly vulnerable to poisoning attacks, where malicious clients could manipulate the system by sending carefully-crafted local models to their neighboring clients. To date, only a limited number of Byzantine-robust DFL methods have been proposed, most of which are either communication-inefficient or remain vulnerable to advanced poisoning attacks. In this paper, we propose a new algorithm called BALANCE (Byzantine-robust averaging through local similarity in decentralization) to defend against poisoning attacks in DFL. In BALANCE, each client leverages its own local model as a similarity reference to determine if the received model is malicious or benign. We establish the theoretical convergence guarantee for BALANCE under poisoning attacks in both strongly convex and non-convex settings. Furthermore, the convergence rate of BALANCE under poisoning attacks matches those of the state-of-the-art counterparts in Byzantine-free settings. Extensive experiments also demonstrate that BALANCE outperforms existing DFL methods and effectively defends against poisoning attacks., Comment: To appear in ACM Conference on Computer and Communications Security 2024 (CCS '24)

We describe the K-moduli spaces of weighted hypersurfaces of degree $2(n+3)$ in $\mathbb{P}(1,2,n+2,n+3)$. We show that the K-polystable limits of these weighted hypersurfaces are also weighted hypersurfaces of the same degree in the same weighted projective space. This is achieved by an explicit study of the wall crossing for K-moduli spaces $M_w$ of certain log Fano pairs with coefficient $w$ whose double cover gives the weighted hypersurface. Moreover, we show that the wall crossing of $M_w$ coincides with variation of GIT except at the last K-moduli wall which gives a divisorial contraction. Our K-moduli spaces provide new birational models for some natural loci in the moduli space of marked hyperelliptic curves., Comment: 52 pages

We introduce Seed-TTS, a family of large-scale autoregressive text-to-speech (TTS) models capable of generating speech that is virtually indistinguishable from human speech. Seed-TTS serves as a foundation model for speech generation and excels in speech in-context learning, achieving performance in speaker similarity and naturalness that matches ground truth human speech in both objective and subjective evaluations. With fine-tuning, we achieve even higher subjective scores across these metrics. Seed-TTS offers superior controllability over various speech attributes such as emotion and is capable of generating highly expressive and diverse speech for speakers in the wild. Furthermore, we propose a self-distillation method for speech factorization, as well as a reinforcement learning approach to enhance model robustness, speaker similarity, and controllability. We additionally present a non-autoregressive (NAR) variant of the Seed-TTS model, named $\text{Seed-TTS}_\text{DiT}$, which utilizes a fully diffusion-based architecture. Unlike previous NAR-based TTS systems, $\text{Seed-TTS}_\text{DiT}$ does not depend on pre-estimated phoneme durations and performs speech generation through end-to-end processing. We demonstrate that this variant achieves comparable performance to the language model-based variant and showcase its effectiveness in speech editing. We encourage readers to listen to demos at \url{https://bytedancespeech.github.io/seedtts_tech_report}.

Simultaneous Machine Translation (SiMT) generates target outputs while receiving stream source inputs and requires a read/write policy to decide whether to wait for the next source token or generate a new target token, whose decisions form a \textit{decision path}. Existing SiMT methods, which learn the policy by exploring various decision paths in training, face inherent limitations. These methods not only fail to precisely optimize the policy due to the inability to accurately assess the individual impact of each decision on SiMT performance, but also cannot sufficiently explore all potential paths because of their vast number. Besides, building decision paths requires unidirectional encoders to simulate streaming source inputs, which impairs the translation quality of SiMT models. To solve these issues, we propose \textbf{S}elf-\textbf{M}odifying \textbf{S}tate \textbf{M}odeling (SM$^2$), a novel training paradigm for SiMT task. Without building decision paths, SM$^2$ individually optimizes decisions at each state during training. To precisely optimize the policy, SM$^2$ introduces Self-Modifying process to independently assess and adjust decisions at each state. For sufficient exploration, SM$^2$ proposes Prefix Sampling to efficiently traverse all potential states. Moreover, SM$^2$ ensures compatibility with bidirectional encoders, thus achieving higher translation quality. Experiments show that SM$^2$ outperforms strong baselines. Furthermore, SM$^2$ allows offline machine translation models to acquire SiMT ability with fine-tuning., Comment: Accept to ACL 2024 main conference. 15 pages, 13 figures, 9 tables

We prove the set of Koll\'{a}r valuations in the dual complex of a klt singularity with a fixed complement is path connected. We also classify the case when the dual complex is one dimensional., Comment: 9 pages, comments welcome. To appear in SCIENCE CHINA Mathematics, special issue in memory of Professor Gang Xiao

While AI-generated content has garnered significant attention, achieving photo-realistic video synthesis remains a formidable challenge. Despite the promising advances in diffusion models for video generation quality, the complex model architecture and substantial computational demands for both training and inference create a significant gap between these models and real-world applications. This paper presents SNED, a superposition network architecture search method for efficient video diffusion model. Our method employs a supernet training paradigm that targets various model cost and resolution options using a weight-sharing method. Moreover, we propose the supernet training sampling warm-up for fast training optimization. To showcase the flexibility of our method, we conduct experiments involving both pixel-space and latent-space video diffusion models. The results demonstrate that our framework consistently produces comparable results across different model options with high efficiency. According to the experiment for the pixel-space video diffusion model, we can achieve consistent video generation results simultaneously across 64 x 64 to 256 x 256 resolutions with a large range of model sizes from 640M to 1.6B number of parameters for pixel-space video diffusion models., Comment: Accepted in CVPR 2024

In the field of autonomous driving, even a meticulously trained model can encounter failures when faced with unfamiliar sceanrios. One of these scenarios can be formulated as an online continual learning (OCL) problem. That is, data come in an online fashion, and models are updated according to these streaming data. Two major OCL challenges are catastrophic forgetting and data imbalance. To address these challenges, in this paper, we propose an Analytic Exemplar-Free Online Continual Learning (AEF-OCL). The AEF-OCL leverages analytic continual learning principles and employs ridge regression as a classifier for features extracted by a large backbone network. It solves the OCL problem by recursively calculating the analytical solution, ensuring an equalization between the continual learning and its joint-learning counterpart, and works without the need to save any used samples (i.e., exemplar-free). Additionally, we introduce a Pseudo-Features Generator (PFG) module that recursively estimates the deviation of real features. The PFG generates offset pseudo-features following a normal distribution, thereby addressing the data imbalance issue. Experimental results demonstrate that despite being an exemplar-free strategy, our method outperforms various methods on the autonomous driving SODA10M dataset. Source code is available at https://github.com/ZHUANGHP/Analytic-continual-learning.

This demo paper examines the susceptibility of Federated Learning (FL) systems to targeted data poisoning attacks, presenting a novel system for visualizing and mitigating such threats. We simulate targeted data poisoning attacks via label flipping and analyze the impact on model performance, employing a five-component system that includes Simulation and Data Generation, Data Collection and Upload, User-friendly Interface, Analysis and Insight, and Advisory System. Observations from three demo modules: label manipulation, attack timing, and malicious attack availability, and two analysis components: utility and analytical behavior of local model updates highlight the risks to system integrity and offer insight into the resilience of FL systems. The demo is available at https://github.com/CathyXueqingZhang/DataPoisoningVis.

We present personalized residuals and localized attention-guided sampling for efficient concept-driven generation using text-to-image diffusion models. Our method first represents concepts by freezing the weights of a pretrained text-conditioned diffusion model and learning low-rank residuals for a small subset of the model's layers. The residual-based approach then directly enables application of our proposed sampling technique, which applies the learned residuals only in areas where the concept is localized via cross-attention and applies the original diffusion weights in all other regions. Localized sampling therefore combines the learned identity of the concept with the existing generative prior of the underlying diffusion model. We show that personalized residuals effectively capture the identity of a concept in ~3 minutes on a single GPU without the use of regularization images and with fewer parameters than previous models, and localized sampling allows using the original model as strong prior for large parts of the image., Comment: CVPR 2024. Project page at https://cusuh.github.io/personalized-residuals

Gushel-Mukai manifolds are specific families of $n$-dimensional Fano manifolds of Picard rank $1$ and index $n-2$ where $3\leq n \leq 6$. A Gushel-Mukai $n$-fold is either ordinary, i.e. a hyperquadric section of a quintic Del Pezzo $(n+1)$-fold, or special, i.e. it admits a double cover over the quintic Del Pezzo $n$-fold branched along an ordinary Gushel-Mukai $(n-1)$-fold. In this paper, we prove that a general special Gushel-Mukai $n$-fold is K-stable for every $3\leq n\leq 6$. Furthermore, we give a description of the first and last walls of the K-moduli of the pair $(M,cQ)$, where $M$ is the quintic Del Pezzo fourfold (or fivefold) and $Q$ is an ordinary Gushel-Mukai threefold (or fourfold). Besides, we compute $\delta$-invariants of quintic Del Pezzo fourfolds and fivefolds which were shown to be K-unstable by K. Fujita, and show that they admit K\"ahler-Ricci solitons., Comment: 31 pages, comments are very welcome

Diffusion Models (DMs) have exhibited superior performance in generating high-quality and diverse images. However, this exceptional performance comes at the cost of expensive architectural design, particularly due to the attention module heavily used in leading models. Existing works mainly adopt a retraining process to enhance DM efficiency. This is computationally expensive and not very scalable. To this end, we introduce the Attention-driven Training-free Efficient Diffusion Model (AT-EDM) framework that leverages attention maps to perform run-time pruning of redundant tokens, without the need for any retraining. Specifically, for single-denoising-step pruning, we develop a novel ranking algorithm, Generalized Weighted Page Rank (G-WPR), to identify redundant tokens, and a similarity-based recovery method to restore tokens for the convolution operation. In addition, we propose a Denoising-Steps-Aware Pruning (DSAP) approach to adjust the pruning budget across different denoising timesteps for better generation quality. Extensive evaluations show that AT-EDM performs favorably against prior art in terms of efficiency (e.g., 38.8% FLOPs saving and up to 1.53x speed-up over Stable Diffusion XL) while maintaining nearly the same FID and CLIP scores as the full model. Project webpage: https://atedm.github.io., Comment: Accepted to IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2024

In recent years, the complexity of 5G and beyond wireless networks has escalated, prompting a need for innovative frameworks to facilitate flexible management and efficient deployment. The concept of digital twins (DTs) has emerged as a solution to enable real-time monitoring, predictive configurations, and decision-making processes. While existing works primarily focus on leveraging DTs to optimize wireless networks, a detailed mapping methodology for creating virtual representations of network infrastructure and properties is still lacking. In this context, we introduce VH-Twin, a novel time-series data-driven framework that effectively maps wireless networks into digital reality. VH-Twin distinguishes itself through complementary vertical twinning (V-twinning) and horizontal twinning (H-twinning) stages, followed by a periodic clustering mechanism used to virtualize network regions based on their distinct geological and wireless characteristics. Specifically, V-twinning exploits distributed learning techniques to initialize a global twin model collaboratively from virtualized network clusters. H-twinning, on the other hand, is implemented with an asynchronous mapping scheme that dynamically updates twin models in response to network or environmental changes. Leveraging real-world wireless traffic data within a cellular wireless network, comprehensive experiments are conducted to verify that VH-Twin can effectively construct, deploy, and maintain network DTs. Parametric analysis also offers insights into how to strike a balance between twinning efficiency and model accuracy at scale., Comment: Accepted by IFIP/IEEE Networking 2024

Federated Learning (FL) offers a distributed framework to train a global control model across multiple base stations without compromising the privacy of their local network data. This makes it ideal for applications like wireless traffic prediction (WTP), which plays a crucial role in optimizing network resources, enabling proactive traffic flow management, and enhancing the reliability of downstream communication-aided applications, such as IoT devices, autonomous vehicles, and industrial automation systems. Despite its promise, the security aspects of FL-based distributed wireless systems, particularly in regression-based WTP problems, remain inadequately investigated. In this paper, we introduce a novel fake traffic injection (FTI) attack, designed to undermine the FL-based WTP system by injecting fabricated traffic distributions with minimal knowledge. We further propose a defense mechanism, termed global-local inconsistency detection (GLID), which strategically removes abnormal model parameters that deviate beyond a specific percentile range estimated through statistical methods in each dimension. Extensive experimental evaluations, performed on real-world wireless traffic datasets, demonstrate that both our attack and defense strategies significantly outperform existing baselines., Comment: Accepted by IFIP/IEEE Networking 2024

The recent breakthroughs in the research on Large Language Models (LLMs) have triggered a transformation across several research domains. Notably, the integration of LLMs has greatly enhanced performance in robot Task And Motion Planning (TAMP). However, previous approaches often neglect the consideration of dynamic environments, i.e., the presence of dynamic objects such as humans. In this paper, we propose a novel approach to address this gap by incorporating human awareness into LLM-based robot task planning. To obtain an effective representation of the dynamic environment, our approach integrates humans' information into a hierarchical scene graph. To ensure the plan's executability, we leverage LLMs to ground the environmental topology and actionable knowledge into formal planning language. Most importantly, we use LLMs to predict future human activities and plan tasks for the robot considering the predictions. Our contribution facilitates the development of integrating human awareness into LLM-driven robot task planning, and paves the way for proactive robot decision-making in dynamic environments.

In the evolving landscape of Neural Machine Translation (NMT), the pretrain-then-finetune paradigm has yielded impressive results. However, the persistent challenge of Catastrophic Forgetting (CF) remains a hurdle. While previous work has introduced Continual Learning (CL) methods to address CF, these approaches grapple with the delicate balance between avoiding forgetting and maintaining system extensibility. To address this, we propose a CL method, named $\textbf{F-MALLOC}$ ($\textbf{F}$eed-forward $\textbf{M}$emory $\textbf{ALLOC}ation)$. F-MALLOC is inspired by recent insights highlighting that feed-forward layers emulate neural memories and encapsulate crucial translation knowledge. It decomposes feed-forward layers into discrete memory cells and allocates these memories to different tasks. By learning to allocate and safeguard these memories, our method effectively alleviates CF while ensuring robust extendability. Besides, we propose a comprehensive assessment protocol for multi-stage CL of NMT systems. Experiments conducted following this new protocol showcase the superior performance of F-MALLOC, evidenced by higher BLEU scores and almost zero forgetting., Comment: Accepted to the main conference of NAACL 2024

Recent advancements in Large Language Models (LLMs) have sparked a revolution across many research fields. In robotics, the integration of common-sense knowledge from LLMs into task and motion planning has drastically advanced the field by unlocking unprecedented levels of context awareness. Despite their vast collection of knowledge, large language models may generate infeasible plans due to hallucinations or missing domain information. To address these challenges and improve plan feasibility and computational efficiency, we introduce DELTA, a novel LLM-informed task planning approach. By using scene graphs as environment representations within LLMs, DELTA achieves rapid generation of precise planning problem descriptions. To enhance planning performance, DELTA decomposes long-term task goals with LLMs into an autoregressive sequence of sub-goals, enabling automated task planners to efficiently solve complex problems. In our extensive evaluation, we show that DELTA enables an efficient and fully automatic task planning pipeline, achieving higher planning success rates and significantly shorter planning times compared to the state of the art.

Federated neuromorphic learning (FedNL) leverages event-driven spiking neural networks and federated learning frameworks to effectively execute intelligent analysis tasks over amounts of distributed low-power devices but also perform vulnerability to poisoning attacks. The threat of backdoor attacks on traditional deep neural networks typically comes from time-invariant data. However, in FedNL, unknown threats may be hidden in time-varying spike signals. In this paper, we start to explore a novel vulnerability of FedNL-based systems with the concept of time division multiplexing, termed Spikewhisper, which allows attackers to evade detection as much as possible, as multiple malicious clients can imperceptibly poison with different triggers at different timeslices. In particular, the stealthiness of Spikewhisper is derived from the time-domain divisibility of global triggers, in which each malicious client pastes only one local trigger to a certain timeslice in the neuromorphic sample, and also the polarity and motion of each local trigger can be configured by attackers. Extensive experiments based on two different neuromorphic datasets demonstrate that the attack success rate of Spikewispher is higher than the temporally centralized attacks. Besides, it is validated that the effect of Spikewispher is sensitive to the trigger duration.

In this article, we study the K-moduli space of Fano threefolds obtained by blowing up $\mathbb{P}^3$ along $(2,3)$-complete intersection curves. This K-moduli space is a two-step birational modification of the GIT moduli space of $(3,3)$-curves on $\mathbb{P}^1 \times \mathbb{P}^1$. As an application, we show that our K-moduli space appears as one model of the Hassett--Keel program for $\overline{M}_4$. In particular, we classify all K-(semi/poly)stable members in this deformation family of Fano varieties. We follow the moduli continuity method with moduli of lattice-polarized K3 surfaces, general elephants and Sarkisov links as new ingredients., Comment: 36 pages, comments are welcome

Online Class Incremental Learning (OCIL) aims to train the model in a task-by-task manner, where data arrive in mini-batches at a time while previous data are not accessible. A significant challenge is known as Catastrophic Forgetting, i.e., loss of the previous knowledge on old data. To address this, replay-based methods show competitive results but invade data privacy, while exemplar-free methods protect data privacy but struggle for accuracy. In this paper, we proposed an exemplar-free approach -- Analytic Online Class Incremental Learning (AOCIL). Instead of back-propagation, we design the Analytic Classifier (AC) updated by recursive least square, cooperating with a frozen backbone. AOCIL simultaneously achieves high accuracy, low resource consumption and data privacy protection. We conduct massive experiments on four existing benchmark datasets, and the results demonstrate the strong capability of handling OCIL scenarios. Codes will be ready.

In this article, the use of channel state information (CSI) for indoor positioning is studied. In the considered model, a server equipped with several antennas sends pilot signals to users, while each user uses the received pilot signals to estimate channel states for user positioning. To this end, we formulate the positioning problem as an optimization problem aiming to minimize the gap between the estimated positions and the ground truth positions of users. To solve this problem, we design a complex-valued neural network (CVNN) model based federated learning (FL) algorithm. Compared to standard real-valued centralized machine learning (ML) methods, our proposed algorithm has two main advantages. First, our proposed algorithm can directly process complex-valued CSI data without data transformation. Second, our proposed algorithm is a distributed ML method that does not require users to send their CSI data to the server. Since the output of our proposed algorithm is complex-valued which consists of the real and imaginary parts, we study the use of the CVNN to implement two learning tasks. First, the proposed algorithm directly outputs the estimated positions of a user. Here, the real and imaginary parts of an output neuron represent the 2D coordinates of the user. Second, the proposed method can output two CSI features (i.e., line-of-sight/non-line-of-sight transmission link classification and time of arrival (TOA) prediction) which can be used in traditional positioning algorithms. Simulation results demonstrate that our designed CVNN based FL can reduce the mean positioning error between the estimated position and the actual position by up to 36%, compared to a RVNN based FL which requires to transform CSI data into real-valued data., Comment: 13 pages, 10 figures

While achieving remarkable progress in a broad range of tasks, large language models (LLMs) remain significantly limited in properly using massive external tools. Existing in-context learning approaches simply format tools into a list of plain text descriptions and input them to LLMs, from which, LLMs generate a sequence of tool calls to solve problems step by step. Such a paradigm ignores the intrinsic dependency between tools and offloads all reasoning loads to LLMs, making them restricted to a limited number of specifically designed tools. It thus remains challenging for LLMs to operate on a library of massive tools, casting a great limitation when confronted with real-world scenarios. This paper proposes ToolNet, a plug-and-play framework that scales up the number of tools to thousands with a moderate increase in token consumption. ToolNet organizes tools into a directed graph. Each node represents a tool, and weighted edges denote tool transition. Starting from an initial tool node, an LLM navigates in the graph by iteratively choosing the next one from its successors until the task is resolved. Extensive experiments show that ToolNet can achieve impressive results in challenging multi-hop tool learning datasets and is resilient to tool failures.

Attosecond photoemission or photoionization delays are a unique probe of the structure and the electronic dynamics of matter. However, spectral congestion and spatial delocalization of valence electron wave functions set fundamental limits to the complexity of systems that can be studied and the information that can be retrieved, respectively. Using attosecond X-ray pulses from LCLS, we demonstrate the key advantages of measuring core-level delays: the photoelectron spectra remain atom-like, the measurements become element specific and the observed scattering dynamics originate from a point-like source. We exploit these unique features to reveal the effects of electronegativity and symmetry on attosecond scattering dynamics by measuring the photoionization delays between N-1s and C-1s core shells of a series of aromatic azabenzene molecules. Remarkably, the delays systematically increase with the number of nitrogen atoms in the molecule and reveal multiple resonances. We identify two previously unknown mechanisms regulating the associated attosecond dynamics, namely the enhanced confinement of the trapped wavefunction with increasing electronegativity of the atoms and the decrease of the coupling strength among the photoemitted partial waves with increasing symmetry. This study demonstrates the unique opportunities opened by measurements of core-level photoionization delays for unravelling attosecond electron dynamics in complex matter.

In this paper, the problem of low-latency communication and computation resource allocation for digital twin (DT) over wireless networks is investigated. In the considered model, multiple physical devices in the physical network (PN) needs to frequently offload the computation task related data to the digital network twin (DNT), which is generated and controlled by the central server. Due to limited energy budget of the physical devices, both computation accuracy and wireless transmission power must be considered during the DT procedure. This joint communication and computation problem is formulated as an optimization problem whose goal is to minimize the overall transmission delay of the system under total PN energy and DNT model accuracy constraints. To solve this problem, an alternating algorithm with iteratively solving device scheduling, power control, and data offloading subproblems. For the device scheduling subproblem, the optimal solution is obtained in closed form through the dual method. For the special case with one physical device, the optimal number of transmission times is reveled. Based on the theoretical findings, the original problem is transformed into a simplified problem and the optimal device scheduling can be found. Numerical results verify that the proposed algorithm can reduce the transmission delay of the system by up to 51.2\% compared to the conventional schemes.

In this paper, the problem of using one active unmanned aerial vehicle (UAV) and four passive UAVs to localize a 3D target UAV in real time is investigated. In the considered model, each passive UAV receives reflection signals from the target UAV, which are initially transmitted by the active UAV. The received reflection signals allow each passive UAV to estimate the signal transmission distance which will be transmitted to a base station (BS) for the estimation of the position of the target UAV. Due to the movement of the target UAV, each active/passive UAV must optimize its trajectory to continuously localize the target UAV. Meanwhile, since the accuracy of the distance estimation depends on the signal-to-noise ratio of the transmission signals, the active UAV must optimize its transmit power. This problem is formulated as an optimization problem whose goal is to jointly optimize the transmit power of the active UAV and trajectories of both active and passive UAVs so as to maximize the target UAV positioning accuracy. To solve this problem, a Z function decomposition based reinforcement learning (ZD-RL) method is proposed. Compared to value function decomposition based RL (VD-RL), the proposed method can find the probability distribution of the sum of future rewards to accurately estimate the expected value of the sum of future rewards thus finding better transmit power of the active UAV and trajectories for both active and passive UAVs and improving target UAV positioning accuracy. Simulation results show that the proposed ZD-RL method can reduce the positioning errors by up to 39.4% and 64.6%, compared to VD-RL and independent deep RL methods, respectively.

Electroencephalogram (EEG) is a non-invasive technique to record bioelectrical signals. Integrating supervised deep learning techniques with EEG signals has recently facilitated automatic analysis across diverse EEG-based tasks. However, the label issues of EEG signals have constrained the development of EEG-based deep models. Obtaining EEG annotations is difficult that requires domain experts to guide collection and labeling, and the variability of EEG signals among different subjects causes significant label shifts. To solve the above challenges, self-supervised learning (SSL) has been proposed to extract representations from unlabeled samples through well-designed pretext tasks. This paper concentrates on integrating SSL frameworks with temporal EEG signals to achieve efficient representation and proposes a systematic review of the SSL for EEG signals. In this paper, 1) we introduce the concept and theory of self-supervised learning and typical SSL frameworks. 2) We provide a comprehensive review of SSL for EEG analysis, including taxonomy, methodology, and technique details of the existing EEG-based SSL frameworks, and discuss the difference between these methods. 3) We investigate the adaptation of the SSL approach to various downstream tasks, including the task description and related benchmark datasets. 4) Finally, we discuss the potential directions for future SSL-EEG research., Comment: 35 pages, 12 figures

In this paper, a novel clustered FL framework that enables distributed edge devices with non-IID data to independently form several clusters in a distributed manner and implement FL training within each cluster is proposed. In particular, our designed clustered FL algorithm must overcome two challenges associated with FL training. First, the server has limited FL training information (i.e., the parameter server can only obtain the FL model information of each device) and limited computational power for finding the differences among a large amount of devices. Second, each device does not have the data information of other devices for device clustering and can only use global FL model parameters received from the server and its data information to determine its cluster identity, which will increase the difficulty of device clustering. To overcome these two challenges, we propose a joint gradient and loss based distributed clustering method in which each device determines its cluster identity considering the gradient similarity and training loss. The proposed clustering method not only considers how a local FL model of one device contributes to each cluster but also the direction of gradient descent thus improving clustering speed. By delegating clustering decisions to edge devices, each device can fully leverage its private data information to determine its own cluster identity, thereby reducing clustering overhead and improving overall clustering performance. Simulation results demonstrate that our proposed clustered FL algorithm can reduce clustering iterations by up to 99% compared to the existing baseline.

Multi-modal Large Language Models (MLLMs) have made significant strides in expanding the capabilities of Large Language Models (LLMs) through the incorporation of visual perception interfaces. Despite the emergence of exciting applications and the availability of diverse instruction tuning data, existing approaches often rely on CLIP or its variants as the visual branch, and merely extract features from the deep layers. However, these methods lack a comprehensive analysis of the visual encoders in MLLMs. In this paper, we conduct an extensive investigation into the effectiveness of different vision encoders within MLLMs. Our findings reveal that the shallow layer features of CLIP offer particular advantages for fine-grained tasks such as grounding and region understanding. Surprisingly, the vision-only model DINO, which is not pretrained with text-image alignment, demonstrates promising performance as a visual branch within MLLMs. By simply equipping it with an MLP layer for alignment, DINO surpasses CLIP in fine-grained related perception tasks. Building upon these observations, we propose a simple yet effective feature merging strategy, named COMM, that integrates CLIP and DINO with Multi-level features Merging, to enhance the visual capabilities of MLLMs. We evaluate COMM through comprehensive experiments on a wide range of benchmarks, including image captioning, visual question answering, visual grounding, and object hallucination. Experimental results demonstrate the superior performance of COMM compared to existing methods, showcasing its enhanced visual capabilities within MLLMs.

We present a study of Lyman continuum (LyC) emission in a sample of $\sim$150 Ly$\alpha$ emitters (LAEs) at $z\approx3.1$ in the Subaru-XMM Deep Survey field. These LAEs were previously selected using the narrowband technique and spectroscopically confirmed with Ly$\alpha$ equivalent widths (EWs) $\ge45$ \r{A}. We obtain deep UV images using a custom intermediate-band filter $U_{\rm J}$ that covers a wavelength range of $3330 \sim 3650$ \r{A}, corresponding to 810$\sim$890 \r{A} in the rest frame. We detect 5 individual LyC galaxy candidates in the $U_{\rm J}$ band, and their escape fractions ($f_{\rm esc}$) of LyC photons are roughly between 40% and 80%. This supports a previous finding that a small fraction of galaxies may have very high $f_{\rm esc}$. We find that the $f_{\rm esc}$ values of the 5 LyC galaxies are not apparently correlated with other galaxy properties such as Ly$\alpha$ luminosity and EW, UV luminosity and slope, and star-formation rate (SFR). This is partly due to the fact that these galaxies only represent a small fraction ($\sim3$%) of our LAE sample. For the remaining LAEs that are not detected in $U_{\rm J}$, we stack their $U_{\rm J}$-band images and constrain their average $f_{\rm esc}$. The upper limit of the average $f_{\rm esc}$ value is about 16%, consistent with the results in the literature. Compared with the non-LyC LAEs, the LyC LAEs tend to have higher Ly$\alpha$ luminosities, Ly$\alpha$ EWs, and SFRs, but their UV continuum slopes are similar to those of other galaxies., Comment: 13 pages, 9 figures. Accepted for publication in ApJ

In contemporary society, voice-controlled devices, such as smartphones and home assistants, have become pervasive due to their advanced capabilities and functionality. The always-on nature of their microphones offers users the convenience of readily accessing these devices. However, recent research and events have revealed that such voice-controlled devices are prone to various forms of malicious attacks, hence making it a growing concern for both users and researchers to safeguard against such attacks. Despite the numerous studies that have investigated adversarial attacks and privacy preservation for images, a conclusive study of this nature has not been conducted for the audio domain. Therefore, this paper aims to examine existing approaches for privacy-preserving and privacy-attacking strategies for audio and speech. To achieve this goal, we classify the attack and defense scenarios into several categories and provide detailed analysis of each approach. We also interpret the dissimilarities between the various approaches, highlight their contributions, and examine their limitations. Our investigation reveals that voice-controlled devices based on neural networks are inherently susceptible to specific types of attacks. Although it is possible to enhance the robustness of such models to certain forms of attack, more sophisticated approaches are required to comprehensively safeguard user privacy.

The emergence of large language models (LLMs) has sparked significant interest in extending their remarkable language capabilities to speech. However, modality alignment between speech and text still remains an open problem. Current solutions can be categorized into two strategies. One is a cascaded approach where outputs (tokens or states) of a separately trained speech recognition system are used as inputs for LLMs, which limits their potential in modeling alignment between speech and text. The other is an end-to-end approach that relies on speech instruction data, which is very difficult to collect in large quantities. In this paper, we address these issues and propose the BLSP approach that Bootstraps Language-Speech Pre-training via behavior alignment of continuation writing. We achieve this by learning a lightweight modality adapter between a frozen speech encoder and an LLM, ensuring that the LLM exhibits the same generation behavior regardless of the modality of input: a speech segment or its transcript. The training process can be divided into two steps. The first step prompts an LLM to generate texts with speech transcripts as prefixes, obtaining text continuations. In the second step, these continuations are used as supervised signals to train the modality adapter in an end-to-end manner. We demonstrate that this straightforward process can extend the capabilities of LLMs to speech, enabling speech recognition, speech translation, spoken language understanding, and speech conversation, even in zero-shot cross-lingual scenarios.

Augmented Language Models (ALMs) empower large language models with the ability to use tools, transforming them into intelligent agents for real-world interactions. However, most existing frameworks for ALMs, to varying degrees, are deficient in the following critical features: flexible customization, collaborative democratization, and holistic evaluation. We present gentopia, an ALM framework enabling flexible customization of agents through simple configurations, seamlessly integrating various language models, task formats, prompting modules, and plugins into a unified paradigm. Furthermore, we establish gentpool, a public platform enabling the registration and sharing of user-customized agents. Agents registered in gentpool are composable such that they can be assembled together for agent collaboration, advancing the democratization of artificial intelligence. To ensure high-quality agents, gentbench, an integral component of gentpool, is designed to thoroughly evaluate user-customized agents across diverse aspects such as safety, robustness, efficiency, etc. We release gentopia on Github and will continuously move forward.

We develop the moduli theory of boundary polarized CY pairs, which are slc Calabi-Yau pairs $(X,D)$ such that $D$ is ample. The motivation for studying this moduli problem is to construct a moduli space at the Calabi-Yau wall interpolating between certain K-moduli and KSBA moduli spaces. We prove that the moduli stack of boundary polarized CY pairs is S-complete, $\Theta$-reductive, and satisfies the existence part of the valuative criterion for properness, which are steps towards constructing a proper moduli space. A key obstacle in this theory is that the irreducible components of the moduli stack are not in general of finite type. Despite this issue, in the case of pairs $(X,D)$ where $X$ is a degeneration of $\mathbb{P}^2$, we construct a projective moduli space on which the Hodge line bundle is ample. As a consequence, we complete the proof of a conjecture of Prokhorov and Shokurov in relative dimension two., Comment: 105 pages, comments are very welcome