| Website | https://ccanonne.github.io/ |
You get the #graph as H, which then allows you to do quick-and-dirty (or slow-and-principled) analysis in Mathematica, should you want to. E.g., generating illustrations for slides :)
For instance, visualise the adjacency matrix with "MatrixPlot[AdjacencyMatrix[H]]"
Here's a simple and quick way to generate a planted clique in an Erdős–Rényi–Gilbert graph in #Mathematica. Below, such a random graph for n=60, k=15, p=1/2.
n = 60; k = 15; p = 1/2;
G = RandomGraph[BernoulliGraphDistribution[n, p]];
K = RandomSample[Range[n], k];
H = EdgeAdd[G, Map[UndirectedEdge[#[[1]], #[[2]]] &, Subsets[K, {2}]], EdgeStyle -> Map[UndirectedEdge[#[[1]], #[[2]]] -> Red &, Subsets[K, {2}]]];
H = Graph[H, VertexStyle -> Map[# -> Red &, K]];
(repost from an earlier tweet)
Considering switching servers on the #fediverse, specifically towards #mathstodon (mathstodon.xyz) whose name seems aligned towards both #maths and #puns, and community seems to share at least some of my interests.
Any experience to share, advice about the pitfalls to avoid and the things to keep in mind? #transferlearning (#notactuallytransferlearning)
A question that came to mind, possibly interesting or trivial: for TV (ℓ₁) there is no dependence on n in the entropy loss. For ℓ∞ a logarithmic dependence pops up.
Does it suddenly appear? Or does it gradually appear when one looks at ℓp guarantees for increasing p? #hashing #computationalcomplexity #theoreticalcomputerscience #theoreticalCompSci #randomness
Jess Wade
