Research on Models Mergence Algorithm Based on Delaunay Triangulation

Fine-tuning pre-trained models for downstream tasks has led to a proliferation of open-sourced task-specific models. Recently, Model Merging (MM) has emerged as an effective approach to facilitate knowledge transfer among these independently fine-tuned models. MM directly combines multiple fine-tuned task-specific models into a merged model without additional training, and the resulting model shows enhanced capabilities in multiple tasks. Although MM provides great utility, it may come with security risks because an adversary can exploit MM to affect multiple downstream tasks. However, the security risks of MM have barely been studied. In this paper, we first find that MM, as a new learning paradigm, introduces unique challenges for existing backdoor attacks due to the merging process. To address these challenges, we introduce BadMerging, the first backdoor attack specifically designed for MM. Notably, BadMerging allows an adversary to compromise the entire merged model by contributing as few as one backdoored task-specific model. BadMerging comprises a two-stage attack mechanism and a novel feature-interpolation-based loss to enhance the robustness of embedded backdoors against the changes of different merging parameters. Considering that a merged model may incorporate tasks from different domains, BadMerging can jointly compromise the tasks provided by the adversary (on-task attack) and other contributors (off-task attack) and solve the corresponding unique challenges with novel attack designs. Extensive experiments show that BadMerging achieves remarkable attacks against various MM algorithms. Our ablation study demonstrates that the proposed attack designs can progressively contribute to the attack performance. Finally, we show that prior defense mechanisms fail to defend against our attacks, highlighting the need for more advanced defense. Our code is available at: https://github.com/jzhang538/BadMerging.

A further generalisation of the planar cubic Bézier spiral

G2 curves composed of planar cubic and Pythagorean hodograph quintic spirals

In Model-Driven Engineering, model merging plays a relevant role in the maintenance and evolution of model-based software. Depending on the amount of metamodels involved in a model merging process, we can classify model merging techniques in two categories: endogenous merging, when all the models to be merged conform to the same metamodel; and exogenous merging, when the models to be merged conform to different metamodels. MOMENT (MOdel manageMENT) is a framework that is integrated in the Eclipse platform, and provides a collection of generic set-oriented operators to manipulate MOF models, following the Model Management discipline. In this paper, we study how model transformations are useful in a model merging process and we provide a solution for both kinds of model merging by means of model management operators and the QVT Relations language.

This paper investigates topology control and seeks to find a distributed solution with low interference, high performance. The paper proposes a distributed topology control algorithm called CBDTG by using the method of adjusting transmission power and the achievements of computational geometry, especially Delaunay triangulation (DT). Each node in the network invokes the neighbor discovery process by using an initial power and adds the discovered neighbors into its neighbor set. Each node finds DT neighbors in order to generate a triangulation with good properties such as planarity and regularity. The experimental result shows that the average node degree is between that of basic CBTC and the condensed CBTC with several edges removed, and the resulting topology will not suffer from high interference due to high node degree, nor will it suffer from low connectivity because of extremely low node degree. CBDTG also outperforms the basic CBTC in throughput.

This paper discusses optimization of quality measures over first order Delaunay triangulations. Unlike most previous work, our measures relate to edge-adjacent or vertex-adjacent triangles instead of only to single triangles. We give efficient algorithms to optimize certain measures, whereas other measures are shown to be NP-hard. For two of the NP-hard maximization problems we provide for any constant ε> 0, factor (1 − ε) approximation algorithms that run in 2O(1/ε)·n and \(2^{O(1/\varepsilon^2)}\cdot n\) time (when the Delaunay triangulation is given). For a third NP-hard problem the NP-hardness proof provides an inapproximability result. Our results are presented for the class of first-order Delaunay triangulations, but also apply to triangulations where every triangle has at most one flippable edge. One of the approximation results is also extended to k-th order Delaunay triangulations.KeywordsDelaunay TriangulationTree DecompositionSteiner PointPolynomial Time Approximation SchemeDelaunay TriangleThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Distribution of uniform reliability over AADTs estimated for a new highway route

Regular triangulations of dynamic sets of points

With the rapid development of the socio-economy and the growing demand for transportation, highways have become an indispensable and important part of the modern transportation network. The design of highway routes and interchanges is a key link in highway construction, which directly affects the efficiency, safety, and traffic smoothness of highways. Based on this, this article analyzes in detail the highway route design and the design of two common interchanges, in order to improve the highway route design and the quality and efficiency of highway construction.

Efficient algorithms for line and curve segment intersection using restricted predicates

Notice of Violation of IEEE Publication Principles<br><br>"Dynamic Removal Algorithm for Constrained Delaunay Triangulations"<br>by Dong-Yu Xu, Guang-Rong Tang, and Yu-Pin Luo<br>in the Proceedings of the Third International Conference on Image and Graphics, 2004,<br><br>After careful and considered review of the content and authorship of this paper by a duly constituted expert committee, this paper has been found to be in violation of IEEE's Publication Principles.<br><br>This paper contains significant portions of original text from the paper cited below. The original text was copied without attribution (including appropriate references to the original author(s) and/or paper title) and without permission.<br><br>Due to the nature of this violation, reasonable effort should be made to remove all past references to this paper, and future references should be made to the following article:<br><br>"Fully Dynamic Constrained Delaunay Triangulations"<br>by Marcelo Kallman, Hanspeter Bieri, and Daniel Thalmann<br>in Geometric Modelling for Scientific Visualization, G. Brunnett, B. Hamann, H.Mueller, L. Linsen (Eds.), Springer-Verlag, 2003, pp. 241-257 <br/> Delaunay triangulations and constrained Delaunay triangulations are very popular tool used to analyze planar domain. An efficient removal algorithm is required when constructing the approximate skeleton of complicated dynamic planar domains. We present a fully dynamic algorithm for CDT. It inserts and removes constraints that maybe points or any kind of polygonal lines. Furthermore, the algorithm automatically detects and deals with the overlapping, self-intersections and duplicate of constraints.

Combining OCL and Programming Languages for UML Model Processing

A conservative particle splitting and merging technique with dynamic pattern and minimum density error

In this paper, we present a flexible partitioning scheme for fractal image compression based on adaptive Delaunay triangulation. Such partition is computed on an initial set of points obtained with a split and merge algorithm in a grey level dependent way. The triangulation is fully flexible and returns a limited number of blocks allowing good compression Ratios Moreover, a vector quantization algorithm is implemented on pixel histograms directly generated from the triangulation. The aim is to reduce the number of comparisons between the two sets of blocks involved in fractal image compression by keeping only the best representative triangles in the domain blocks set. Quality coding results are achieved at rates between 0.06 b/pixel and 0.2 b/pixel for a PSNR between 22 and 25 dB depending on the nature of the original image and on the number of triangles refereed.

Decision tables are a useful technique for implementing complex decision logic, and a concise communication device. Model merging is a process that can greatly benefit from the flexibility of control, rapidity of change, and understandability of purpose that tables engender. Heretofore, users have been cut off from the inner workings of merge tools. Among its many features, our model merging tool, Mirador, opens the process of merging to inspection and manipulation. The tool's support for user modification of the decision table rules that drive its conflict detection and resolution, as well as the possibility of adding customized table conditions and actions is the focus of this paper.