Copy of FYP presentation

IH
by Isaac Harris2 views·Updated Nov 13, 2020
Create your own

Audio TranscriptAuto-generated

  1. graphical representations of probabilistic models.
  2. A variety of real life systems exhibit probabilistic behavior, including
  3. things like security protocols and algorithms that rely on a
  4. degree of randomness to function.
  5. We usually want to verify these systems behaviors we think
  6. they do, and so use probabilistic model checkers to analyze
  7. them. One of these tools is Prism, which allows models
  8. to be described in its own language and then analyze
  9. the model and infer things about the way it functions.
  10. Prism is able to explore a variety of models, from
  11. simple deterministic ones to more complicated ones where each action
  12. has a probability to succeed or fail.
  13. The current problem is that Prism allows models to be
  14. explored to a simulator.
  15. This lets the user step for the states in the
  16. model, choosing which transaction they won't take.
  17. However, there is no way to see this visually.
  18. This makes it much harder to understand the relationships between
  19. states on the overall structure of the model itself.
  20. Currently, models could be explored is a dot file and
  21. viewed using graph is a third party piece of software
  22. used to visualize graphs.
  23. This approaches many problems.
  24. Users must export the model into a dot file every
  25. time they want to view any changes.
  26. They're also unable to explore the graph in real time
  27. to get a better idea of hurricane looks, as prison
  28. currently only exports the entire model.
  29. If the user is only interested in a small segment
  30. of a larger model, they are also unable to visualize
  31. this by itself.
  32. My project aims to fix all of these problems by
  33. allowing a model to be visualized and explored in real
  34. time. I'm creating a G Y that will allow users
  35. to view a graphical representation of the bottle were frontier
  36. of unexplored states.
  37. They can then choose which states they want to explore
  38. on. The graph will reformat itself to display the new
  39. state. As user expands the states, the grass lake will
  40. have to change in such a way that the user
  41. does not get lost.
  42. This will allow users to quickly explore and understand the
  43. model A ZX well as to only view the areas
  44. they need.
  45. Users will also be able to fully explore the graph
  46. or explore it for certain death as required.
  47. The project's main aim is to display as much information
  48. about the model to the user in it's clear away
  49. as possible.
  50. Prism currently allows users to see a path showing what
  51. states they visited on what states they're currently in.
  52. Alternatively, this could be displayed graphically, as shown below, which
  53. is much clearer to understand what states have been explored
  54. on what states connected to those states to decide on
  55. the best solution for drawing the model as a graph,
  56. I first researched what makes a graph readable purchase.
  57. It l studied 49 subjects, looking at a variety of
  58. dense graphs and found it was important to minimize the
  59. number of edge crossings a ZX well as a number
  60. of Ben's in the edges.
  61. The found mixed evidence that suggested the importance of symmetry.
  62. Because of this, I decided to make it a priority
  63. that my algorithm would minimize edge crossings and Ben's.
  64. This led me to a type of graph during algorithm
  65. called a force directed to algorithm, which draws graphs that
  66. tend to be aesthetically pleasing, exhibit symmetries and tend to
  67. produce crossing free layouts for playing our graphs.
  68. My project also to present the interesting problem that the
  69. graph must be redrawn in real time without the use
  70. of getting confused about where they're looking, which force directed
  71. algorithms are able to do what's directed algorithms work by.
  72. Applying force is on the nodes, and then treating the
  73. system is a physical simulation.
  74. We know it's pushing each other apart until an equilibrium
  75. is reached and the nodes stop moving.
  76. When I began this project, I had to decide whether
  77. to build my tool is part of the pre existing
  78. prison code base or to make it a separate tool
  79. that uses the prison.
  80. AP. I made the decision to build.
  81. It, is a separate door first and then potentially integrate
  82. the code into prison itself later.
  83. This then made the next decision.
  84. I had to make much simpler, which Java Graphics Library
  85. would I use for my G as prism is written
  86. in swing, it made sense to also use swing, so
  87. I would be able to easily add the functionality directly
  88. into prison.
  89. To my trace of software engineering methodology, I decided on
  90. the agile methodology for several reasons.
  91. Firstly, it allows me to change the requirements as I
  92. explore the problem and discover new features that may be
  93. useful to add.
  94. This includes things like the extra directions I could take
  95. the project Inputs based visualization parties built a swell as
  96. potentially integrating it into prison directly due to the length
  97. of the project.
  98. It will also allow me to spend more time researching
  99. and programming unless time creating documentation diagrams.
  100. The project has a great amount of room to expand
  101. into as I make progress of it.
  102. Some of these areas include adding the ability to approximate
  103. a heat map of the model.
  104. This would add a way of finding out the most
  105. likely states of the model to be in without having
  106. to perform the complex calculations used to find the exact
  107. answer. There is also the possibility of adding support policies
  108. which will show the best way to reach a particular
  109. state. Thank you for listening to this video, and I
  110. hope you enjoy this presentation.