Scott H. Hawley

How do you “do-”?

Learning “soft-separable” directions in latent space as an alternative to explicit factorization.

Classification as Pattern Recognition

Adding Inference Controls to Pretrained (Latent) Flow Models

Basic physics provides a “straight, fast” way to get up to speed with flow-based generative models

Toy Transformer Tunes

If you can get this one thing, the rest will make sense.

An exploration of the heart of neural audio codecs

Now that we understand contrastive losses and what they’re good for – let’s get rid of them!

Given enough dimensions, does it really matter which distance norm you use?

Depends on whether you’re using unit vectors or not.

Clifford algebra isn’t so scary

Once I figure out a complicated command, I make a bash function for it

Triplet Loss kicks butt

With my new data-visualization Callback!

Visualizations in 3 (and 4) dimensions

In which we test GPT-3’s ability to generate members of a category

A snarky interactive lesson on classification and surveillance

Supporting scholarly blogging by drawing references from your database.

Mac and I aren’t breaking up, we’re just seeing other people.