Recent foundation models are capable of handling multiple tasks and multiple data modalities with the unified base model structure and several specialized model components. However, efficient training of such multi-task (MT) multi-modal (MM) models poses significant system challenges due to the sophisticated model architecture and the heterogeneous workloads of different tasks and modalities. In this paper, we propose Spindle, a brand new training system tailored for resource-efficient and high-performance training of MT MM models via wavefront scheduling. The key idea of Spindle is to decompose the model execution into waves and address the joint optimization problem sequentially, including both heterogeneity-aware workload parallelization and dependency-driven execution scheduling. We build our system and evaluate it on various MT MM models. Experiments demonstrate the superior performance and efficiency of Spindle, with speedup ratio up to 71% compared to state-of-the-art training systems.

In this work, we propose a novel approach of volume visualization without explicit traditional rendering pipeline. In our proposed method, volumetric images can be interactively `reversed' given the volumetric data and a static volume rendered image under the desired rendering effect. Our pipeline enables 3D-navigation on it for exploring the given volumetric data without explicit transfer function. In our approach, deep neural networks, combined usage of Generative Adversarial Networks (GANs) and Convolutional Neural Networks (CNN) are employed to synthesize high-resolution and perceptually authentic images directly, inheriting the desired transfer function and viewing parameter implicitly given by the input images respectively.

In this work, we present a novel access pattern estimation approach for parallel particle tracing in flow field visualization based on deep neural networks. With strong generalization ability, we develop a Long Short-term Memory (LSTM)-based model, which is capable of learning accurate access patterns with only a few training samples and representing the learned patterns with small storage overhead. Equipped with prediction and prefetching functions driven by the developed model, our parallel particle tracing framework employs CPUs and GPUs together for particle tracing tasks. We demonstrate the accuracy and time efficiency of our approach with various flow visualization applications in three different flow datasets.

We propose a dynamically load-balanced algorithm for parallel particle tracing, which periodically attempts to evenly redistribute particles across processes based on k-d tree decomposition. Each process is assigned with (1) a statically partitioned, axis-aligned data block that partially overlaps with neighboring blocks in other processes and (2) a dynamically determined k-d tree leaf node that bounds the active particles for computation; the bounds of the k-d tree nodes are constrained by the geometries of data blocks. Given a certain degree of overlap between blocks, our method can balance the number of particles as much as possible. Compared with other load-balancing algorithms for parallel particle tracing, the proposed method does not require any preanalysis, does not use any heuristics based on flow features, does not make any assumptions about seed distribution, does not move any data blocks during the run, and does not need any master process for work redistribution. Based on a comprehensive performance study up to 8K processes on a Blue Gene/Q system, the proposed algorithm outperforms baseline approaches in both load balance and scalability on various flow visualization and analysis problems.

In this paper, we present a novel scalable approach for visualizing multivariate unsteady flow data with Lagrangian-based Attribute Space Projection (LASP). The distances between spatial temporal samples are evaluated by their attribute values along the advection directions in the flow field. The massive samples are then projected into 2D screen space for feature identification and selection. A hybrid parallel system, which tightly integrates a MapReduce-style particle tracer with a scalable algorithm for massive projection, is designed to support the large scale analysis. Results show that the proposed methods and system are capable of visualizing features in the unsteady flow, which couples multivariate analysis of vector and scalar attributes with projection.

In this paper, we present a novel feature extraction approach called FLDA for unsteady flow fields based on Latent Dirichlet allocation (LDA) model. Analogous to topic modeling in text analysis, in our approach, pathlines and features in a given flow field are defined as documents and words respectively. Flow topics are then extracted based on Latent Dirichlet allocation. Different from other feature extraction methods, our approach clusters pathlines with probabilistic assignment, and aggregates features to meaningful topics at the same time. We build a prototype system to support exploration of unsteady flow field with our proposed LDA-based method. Interactive techniques are also developed to explore the extracted topics and to gain insight from the data. We conduct case studies to demonstrate the effectiveness of our proposed approach.

The secure interaction between two or more administrative domains is a major concern. IRBAC2000 is a model that quickly establishes a flexible policy for dynamic role translation from foreign domains to local. A-IRBAC2000 mode utilizes RBAC to manage dynamic role translation between foreign and local domains. We will see that these mechanisms have significant shortcomings. We propose an improved administrative usage control model named AUCON to overcome the weakness of previous models. AUCON provides administrates user-role assignment for local and foreign domain with unified method. It provides flexible enough mechanism to distinguish users of foreign and local domain and can enforce more strict control for foreign user. While retaining the advantage of traditional RBAC model, AUCON model is being implemented in experiment system

Web service is a new service-oriented computing paradigm which poses the unique security challenges due to its inherent heterogeneity, multi-domain characteristic and highly dynamic nature. A key challenge in Web services security is the design of effective access control schemes. However, most current access control systems base authorization decisions on subject's identity. Administrative scalability and control granularity are serious problems in those systems, and they are not fit for Web services environment. So an attribute-based access control model (WS-ABAC) is presented to address these issues in this paper. WS-ABAC grants access to services based on attributes of the related entities, and uses automated trust negotiation mechanism to address the disclosure issue of the sensitive attributes. It can provide administratively scalable alternative to identity-based authorization methods and provide fine-grained access control for Web services. Moreover, it also can protect user's privacy.

A new fair e-cash payment scheme based on credit is present in this paper. In the scheme, an overdraft credit certificate is issued to user by bank. Using the overdraft credit certificate, user can produce e-cash himself to pay in exchanges. Merchant can verify the e-cash received from user. Bank can make a fair dispute resolution when there is a dissension between user and merchant. It can avoid the problem of partition e-cash for changes, prevent from reusing e-cash and faking e-cash. It fits justice, anonymity, non-deny and impartiality.

Mobile ad hoc networks (MANET) is a new networking paradigm for wireless hosts. Because of infrastructureless, self-organization, dynamic topology and openness of wireless links, the routing security problem in MANET is more seriously than in wired networks. Optimized link state routing (OLSR) (T. Clausen et al., 2003) is proposed by IETF's MANET Group at 2003. In OLSR, neighbor detection is not invulnerable when two bad nodes perform wormhole attack. Furthermore, OLSR's security cannot simply rely on IPSec, because OLSR's packets are often broadcasted and IPsec provides end-to-end security. In this paper, we propose a solution to secure OLSR, which apply the wormhole detective mechanism and authentication to strengthen the neighbor relationship establishment, and use hash-chain and digital signature to protect the routing packets.

Trust management system has been a promising approach to solve the access control problems in distributed systems. Delegation is a core concept in it and needs to be limited with respect to depth. In this paper, some different delegation depth control approaches in current trust management system are discussed. Then RT+/sub 0/ is introduced, which incorporates the integer delegation depth control into RT/sub 0/ The RT+/sub 0/ credential adds to RT/sub 0/ depth value, which provides a more expressive power. The changed semantics is formally defined by a translation from credential to datalog rules. The computational complexity analysis is given and it shows that the semantics is also algorithmically tractable.

Practical adaptive neural control is presented for a class of nonlinear systems with unknown time delays in strict-feedback form. Using appropriate Lyapunov-Krasovskii functionals, the unknown time delays are compensated for. Controller singularity problems are solved by practical neural network control. A novel differentiable control function is provided such that the practical design can be carried out in the decoupled backstepping design. It is proved that the proposed design method is able to guarantee semi-global uniform ultimate boundedness of all the signals in the closed-loop system, and the tracking error is proven to converge to a small neighborhood of the origin.

In this paper, adaptive neural control is presented for a class of strict-feedback nonlinear systems with unknown time delays. The proposed design method does not require a priori knowledge of the signs of the unknown virtual control coefficients. The unknown time delays are compensated for using appropriate Lyapunov-Krasovskii functionals in the design. It is proved that the proposed backstepping design method is able to guarantee semiglobal uniformly ultimately boundedness of all the signals in the closed-loop. In addition, the output of the system is proven to converge to a small neighborhood of the origin. Simulation results are provided to show the effectiveness of the proposed approach.

A class of energy-based position controllers for a kind of flexible spacecraft is proposed. Closed-loop stability of the original distributed parameter system can be achieved, as well as asymptotic stability for the truncated system, which is obtained through representing the deflection of the appendage by an arbitrary finite number of flexible modes. The feedback gains of the controller are tuned by a genetic algorithm (GA) optimization process to achieve good results for tip motion based on some suitable fitness functions. Numerical simulations are carried out on a kind of spacecraft with one flexible appendage, and satisfactory results are obtained.