Denoising Diffusion Probabilistic Models (DDPMs) have accomplished much in the realm of generative AI. Despite their high performance, there is room for improvement, especially in terms of sample fidelity by utilizing statistical properties that impose structural integrity, such as isotropy. Minimizing the mean squared error between the additive and predicted noise alone does not impose constraints on the predicted noise to be isotropic. Thus, we were motivated to utilize the isotropy of the additive noise as a constraint on the objective function to enhance the fidelity of DDPMs. Our approach is simple and can be applied to any DDPM variant. We validate our approach by presenting experiments conducted on four synthetic 2D datasets as well as on unconditional image generation. As demonstrated by the results, the incorporation of this constraint improves the fidelity metrics, Precision and Density for the 2D datasets as well as for the unconditional image generation.
相關內容
Large Language Models (LLMs) have profoundly changed the world. Their self-attention mechanism is the key to the success of transformers in LLMs. However, the quadratic computational cost $O(n^2)$ to the length $n$ input sequence is the notorious obstacle for further improvement and scalability in the longer context. In this work, we leverage the convolution-like structure of attention matrices to develop an efficient approximation method for attention computation using convolution matrices. We propose a $\mathsf{conv}$ basis system, "similar" to the rank basis, and show that any lower triangular (attention) matrix can always be decomposed as a sum of $k$ structured convolution matrices in this basis system. We then design an algorithm to quickly decompose the attention matrix into $k$ convolution matrices. Thanks to Fast Fourier Transforms (FFT), the attention {\it inference} can be computed in $O(knd \log n)$ time, where $d$ is the hidden dimension. In practice, we have $ d \ll n$, i.e., $d=3,072$ and $n=1,000,000$ for Gemma. Thus, when $kd = n^{o(1)}$, our algorithm achieve almost linear time, i.e., $n^{1+o(1)}$. Furthermore, the attention {\it training forward} and {\it backward gradient} can be computed in $n^{1+o(1)}$ as well. Our approach can avoid explicitly computing the $n \times n$ attention matrix, which may largely alleviate the quadratic computational complexity. Furthermore, our algorithm works on any input matrices. This work provides a new paradigm for accelerating attention computation in transformers to enable their application to longer contexts.
Large Multimodal Models (LMMs) have achieved impressive success in visual understanding and reasoning, remarkably improving the performance of mathematical reasoning in a visual context. Yet, a challenging type of visual math lies in the multimodal graph theory problem, which demands that LMMs understand the graphical structures accurately and perform multi-step reasoning on the visual graph. Additionally, exploring multimodal graph theory problems will lead to more effective strategies in fields like biology, transportation, and robotics planning. To step forward in this direction, we are the first to design a benchmark named VisionGraph, used to explore the capabilities of advanced LMMs in solving multimodal graph theory problems. It encompasses eight complex graph problem tasks, from connectivity to shortest path problems. Subsequently, we present a Description-Program-Reasoning (DPR) chain to enhance the logical accuracy of reasoning processes through graphical structure description generation and algorithm-aware multi-step reasoning. Our extensive study shows that 1) GPT-4V outperforms Gemini Pro in multi-step graph reasoning; 2) All LMMs exhibit inferior perception accuracy for graphical structures, whether in zero/few-shot settings or with supervised fine-tuning (SFT), which further affects problem-solving performance; 3) DPR significantly improves the multi-step graph reasoning capabilities of LMMs and the GPT-4V (DPR) agent achieves SOTA performance.
Unmanned Aerial Systems (UASs) or drones become more and more commercially available and cheap. There has been much emphasis on developing and deploying Counter-UAS systems (UASs) with Detection Tracking and Identification (DTI) solutions. However, the capabilities of these systems are hard to benchmark. Performance claims of these systems are currently not supported by evidence. In addition, no standard test methodologies are available for these DTI systems and different test methodologies make comparison of these systems hard or impossible. We report on the definition, development and verification of an objective-driven test method and corresponding comparative performance evaluation for commercial DTI solutions for C-UASs. The developed methodology is based on end-user scenarios that are operationally relevant. The test methodology is based on a generic DTI system lay-out and is detailed towards detection, tracking and identification, taking into account contextual information and end-user input. The comparative performance evaluation is developed to enable the use of the methodology in a relevant environment, thereby taking into account any potential environmental aspect that might influence DTI system performance. Validation of the work in a relevant environment has been done in three operational trials. The operational trial results show that the method allows for performance evaluation at component level (i.e., detection, tracking or identification component) and at system level (combinations of these components and integrated DTI system of system solutions).
Recent High-Performance Computing (HPC) systems are facing important challenges, such as massive power consumption, while at the same time significantly under-utilized system resources. Given the power consumption trends, future systems will be deployed in an over-provisioned manner where more resources are installed than they can afford to power simultaneously. In such a scenario, maximizing resource utilization and energy efficiency, while keeping a given power constraint, is pivotal. Driven by this observation, in this position paper we first highlight the recent trends of resource management techniques, with a particular focus on malleability support (i.e., dynamically scaling resource allocations/requirements for a job), co-scheduling (i.e., co-locating multiple jobs within a node), and power management. Second, we consider putting them together, assess their relationships/synergies, and discuss the functionality requirements in each software component for future over-provisioned and power-constrained HPC systems. Third, we briefly introduce our ongoing efforts on the integration of software tools, which will ultimately lead to the convergence of malleability and power management, as it is designed in the HPC PowerStack initiative.
Autonomous Vehicles (AVs) are prone to revolutionise the transportation industry. However, they must be thoroughly tested to avoid safety violations. Simulation testing plays a crucial role in finding safety violations of Automated Driving Systems (ADSs). This paper proposes PAFOT, a position-based approach testing framework, which generates adversarial driving scenarios to expose safety violations of ADSs. We introduce a 9-position grid which is virtually drawn around the Ego Vehicle (EV) and modify the driving behaviours of Non-Playable Characters (NPCs) to move within this grid. PAFOT utilises a single-objective genetic algorithm to search for adversarial test scenarios. We demonstrate PAFOT on a well-known high-fidelity simulator, CARLA. The experimental results show that PAFOT can effectively generate safety-critical scenarios to crash ADSs and is able to find collisions in a short simulation time. Furthermore, it outperforms other search-based testing techniques by finding more safety-critical scenarios under the same driving conditions within less effective simulation time.
The emerging Self-Sovereign Identity (SSI) techniques, such as Decentralized Identifiers (DIDs) and Verifiable Credentials (VCs), move control of digital identity from conventional identity providers to individuals and lay down the foundation for people, organizations, and things establishing rich digital relationship. The existing applications of SSI mainly focus on creating person-to-person and person-to-service relationships, whereas person-to-device and device-to-device interactions have been largely overlooked. In this paper, we close this gap by identifying a number of key challenges of applying SSI to the Internet of Things (IoT) and providing a comprehensive taxonomy and usage of VCs in the IoT context with respect to their validity period, trust and interoperability level, and scope of usage. The life-cycle management of VCs as well as various optimization techniques for realizing SSI in IoT environments are also addressed in great detail. This work is a noteworthy step towards massive adoption of SSI for securing existing and future IoT applications in practice.
The Digital Services Act (DSA) is a much awaited platforms liability reform in the European Union that was adopted on 1 November 2022 with the ambition to set a global example in terms of accountability and transparency. Among other obligations, the DSA emphasizes the need for online platforms to report on their content moderation decisions (`statements of reasons' - SoRs), which is a novel transparency mechanism we refer to as automated transparency in this study. SoRs are currently made available in the DSA Transparency Database, launched by the European Commission in September 2023. The DSA Transparency Database marks a historical achievement in platform governance, and allows investigations about the actual transparency gains, both at structure level as well as at the level of platform compliance. This study aims to understand whether the Transparency Database helps the DSA to live up to its transparency promises. We use legal and empirical arguments to show that while there are some transparency gains, compliance remains problematic, as the current database structure allows for a lot of discretion from platforms in terms of transparency practices. In our empirical study, we analyze a representative sample of the Transparency Database (131m SoRs) submitted in November 2023, to characterise and evaluate platform content moderation practices.
Owing to their powerful semantic reasoning capabilities, Large Language Models (LLMs) have been effectively utilized as recommenders, achieving impressive performance. However, the high inference latency of LLMs significantly restricts their practical deployment. To address this issue, this work investigates knowledge distillation from cumbersome LLM-based recommendation models to lightweight conventional sequential models. It encounters three challenges: 1) the teacher's knowledge may not always be reliable; 2) the capacity gap between the teacher and student makes it difficult for the student to assimilate the teacher's knowledge; 3) divergence in semantic space poses a challenge to distill the knowledge from embeddings. To tackle these challenges, this work proposes a novel distillation strategy, DLLM2Rec, specifically tailored for knowledge distillation from LLM-based recommendation models to conventional sequential models. DLLM2Rec comprises: 1) Importance-aware ranking distillation, which filters reliable and student-friendly knowledge by weighting instances according to teacher confidence and student-teacher consistency; 2) Collaborative embedding distillation integrates knowledge from teacher embeddings with collaborative signals mined from the data. Extensive experiments demonstrate the effectiveness of the proposed DLLM2Rec, boosting three typical sequential models with an average improvement of 47.97%, even enabling them to surpass LLM-based recommenders in some cases.
Knowledge Graph Embedding (KGE) aims to learn representations for entities and relations. Most KGE models have gained great success, especially on extrapolation scenarios. Specifically, given an unseen triple (h, r, t), a trained model can still correctly predict t from (h, r, ?), or h from (?, r, t), such extrapolation ability is impressive. However, most existing KGE works focus on the design of delicate triple modeling function, which mainly tells us how to measure the plausibility of observed triples, but offers limited explanation of why the methods can extrapolate to unseen data, and what are the important factors to help KGE extrapolate. Therefore in this work, we attempt to study the KGE extrapolation of two problems: 1. How does KGE extrapolate to unseen data? 2. How to design the KGE model with better extrapolation ability? For the problem 1, we first discuss the impact factors for extrapolation and from relation, entity and triple level respectively, propose three Semantic Evidences (SEs), which can be observed from train set and provide important semantic information for extrapolation. Then we verify the effectiveness of SEs through extensive experiments on several typical KGE methods. For the problem 2, to make better use of the three levels of SE, we propose a novel GNN-based KGE model, called Semantic Evidence aware Graph Neural Network (SE-GNN). In SE-GNN, each level of SE is modeled explicitly by the corresponding neighbor pattern, and merged sufficiently by the multi-layer aggregation, which contributes to obtaining more extrapolative knowledge representation. Finally, through extensive experiments on FB15k-237 and WN18RR datasets, we show that SE-GNN achieves state-of-the-art performance on Knowledge Graph Completion task and performs a better extrapolation ability.
Graph Neural Networks (GNNs) have been studied from the lens of expressive power and generalization. However, their optimization properties are less well understood. We take the first step towards analyzing GNN training by studying the gradient dynamics of GNNs. First, we analyze linearized GNNs and prove that despite the non-convexity of training, convergence to a global minimum at a linear rate is guaranteed under mild assumptions that we validate on real-world graphs. Second, we study what may affect the GNNs' training speed. Our results show that the training of GNNs is implicitly accelerated by skip connections, more depth, and/or a good label distribution. Empirical results confirm that our theoretical results for linearized GNNs align with the training behavior of nonlinear GNNs. Our results provide the first theoretical support for the success of GNNs with skip connections in terms of optimization, and suggest that deep GNNs with skip connections would be promising in practice.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191