https://github.com/geyang/sparse-graph

https://github.com/geyang/sparse-graph

graph-search graphs sparse-graphs

Last synced: about 1 month ago
JSON representation

• Host: GitHub
• URL: https://github.com/geyang/sparse-graph
• Owner: geyang
• License: mit
• Created: 2020-10-28T06:18:04.000Z (about 4 years ago)
• Default Branch: main
• Last Pushed: 2021-06-24T20:16:50.000Z (over 3 years ago)
• Last Synced: 2024-04-14T15:11:53.194Z (8 months ago)
• Topics: graph-search, graphs, sparse-graphs
• Language: Jupyter Notebook
• Homepage:
• Size: 3.46 MB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README
  • License: LICENSE

Awesome Lists containing this project

README

        
Dynamic Graphical Memory Implementation

=======================================

-  `planning.py `__ shows how to use this graph

   with graph_search.

-  `graph_sparsification.py `__

   shows how to sparsify a graph

-  `batch_sparsification.py `__

   efficiently sparsify a graph with as a batch

-  `incremental_api.py `__ shows how to

   extend with sparsity

Usage

-----

To instantiate a graph:

.. code-block:: python

   graph = AsymMesh(n=10_000, k=6, dim=2, img_dim=[2], kernel_fn=l2, embed_fn=id2D, d_max=20)

   graph.extend(xys, images=xys, meta=xys)

   graph.update_zs()

   graph.update_edges()

Most of the time it is better to enforce sparsity of the graph by only

adding new vertices when there is no existing vertex that is close-by.

The new dedupe API allows us to do this in a batch fashion:

.. code-block:: python

   spots = graph.dedupe(images=xys, r_min=r_min)

   xys = xys[spots]

   ds = graph.to_goal(zs_2=xys)

   if ds.size == 0:

       graph.extend(xys, images=xys, meta=xys)

   else:

       m = ds.min(axis=-1) >= r_min

       if m.sum() > 0:

           graph.extend(xys[m], images=xys[m], meta=xys[m])

       graph.update_edges()

+---------------------+--------------+--------------------------------+

| Dense               | Sparse       | Details                        |

+=====================+==============+================================+

| |dense_graph|       | |image1|     | 10x more edges in the dense    |

|                     |              | graph in comparison to the     |

|                     |              | sparse graph.                  |

+---------------------+--------------+--------------------------------+

To Experiment

-------------

Import this module in the method module.

.. |dense_graph| image:: figures/random_graph.png?raw=true

.. |image1| image:: figures/batch_graph.png?raw=true