549 Followers
196 Following
242 Posts

Distinguished Scientist working at the intersection of machine learning and technology infrastructure such as distributed systems, observability, control, time-series, and others.

Things I like :

#Mathematics
#ArtificialIntelligence
#AI
#MachineLearning
#ML
#Causality
#Networks
#Philosophy
#Algorithms

Twitterhttps://twitter.com/EdHenry_
GitHubhttps://github.com/edhenry
Bloghttps://edhenry.github.io
Great lunar halo tonight! Phone camera doesn't do it justice.
And then it took a bite.
Split screen functionality on the phone to the rescue!

Very much enjoyed this post on "typicality" as described by @sedielem and how our own intuitions about high dimensional space may not align with reality.

The image shows how the probability of the number of heads in a sequence of unfair coin tosses changes accounting for ordering of sequences of events rather than just the statistics of event occurrences.

I also like gentle nudge to the OOD field to test something like typicality as opposed to thresholding likelihoods.

https://benanne.github.io/2020/09/01/typicality.html

Musings on typicality

A summary of my current thoughts on typicality, and its relevance to likelihood-based generative models.

Sander Dieleman

I've been using Tusky for Android for a few weeks now. I've recently switched to the Mastodon client and I'm missing the functionality to hide stats related to people, and toots.

I've found myself glancing back at the stats again in the Mastodon client, as I can't find a way to hide them, and I don't like it.

Given the news this morning, I wonder how long this request will actually take. I've been waiting about 30 - 40 minutes now.

I requested my Facebook data at the same time and had it within about 15-20 minutes.

Probably my favorite feature of Mastodon (or Tusky) so far