Innvariant Tools

Provides some additional functionalities re-used over multiple experiments and tools of our group.

Install

• pip install innvariant
• conda:

dependencies:
- pip
- pip:
  - innvariant

• poetry install innvariant[all]

Caching Tools

A simple caching mechanism for computations in jupyter notebooks. The cache is stored locally in disk such that you have some overhead in writing it. Otherwise, better use lru_cache! This solution is intended for cases where you might expect cells to crash etc.

import numpy as np
from innvariant.tools import cache

@cache(key="complex_computation")
def complex_computation(param1: int = 100):
    return {ix: np.random.randn(np.random.randint(10, 20)) for ix in range(param1)}

complex_computation()  # executed first time
complex_computation()  # should be coming from cache

With S3FS support:

import numpy as np
from innvariant.tools import cache

@cache(key="another_computation", s3_access_key="my-access-key", s3_secret_key="my-secret-key", s3_base="/bucket/cache/")
def another_computation(param1: int = 100):
    return {ix: np.random.randn(np.random.randint(10, 20)) for ix in range(param1)}

another_computation()  # executed first time
another_computation()  # should be coming from cache
# Will sync cache to remote s3fs

NetworkX Additions

Generate a glock graph (see @riis2006information):

from innvariant.networkx.generators import glock_graph

glock_graph(n=5, r=2)  # Glock Graph with guessing number r=2

References

• Information flows, graphs and their guessing numbers

@inproceedings{riis2006information,
  title={Information flows, graphs and their guessing numbers},
  author={Riis, S{\o}ren},
  booktitle={2006 4th International Symposium on Modeling and Optimization in Mobile, Ad Hoc and Wireless Networks},
  pages={1--9},
  year={2006},
  organization={IEEE}
}