Allison Parrish @aparrish@friend.camp

the footnote conventions in the translation of the new testament i'm reading led to this particularly meme-ey formulation

I haven't even played the goose game so maybe this is explicitly referenced in there, but still, wow (from Armstrong, Edward A. “The Symbolism of the Swan and the Goose.” Folklore, vol. 55, no. 2, 1944, pp. 54–58. JSTOR, www.jstor.org/stable/1256902.)

a very simplified rendition of smorgasburg (the open air food market in prospect park)

playing around in bitsy, trying to make a game about prospect park

the umap plots for BPEmb's multi-language embeddings look like someone literally spilled some language on the floor https://nlp.h-its.org/bpemb/multi/

using that variational autoencoder I trained on 1m lines of poetry—making a poetic form I'm going to call the "variational sestinoid," which forms its stanzas from samples in progressively larger circles centered on the same point in the latent space

oh hey I found the tutorial level in the New Testament (Acts 12:5–10), in which the angel shows Peter how to equip armor and goes through the basic stealth mechanics. I guess the scribes left out the part where Peter dinged levels two and three from killing dungeon rats on his way out

did this hike this morning https://hikingguy.com/hiking-trails/best-la-hikes/runyon-canyon-hike-directions/ and it was very nice! there were a lot of steep descents that were tough for me (on account of my knee which even years after surgery is still not 100%). i enjoyed saying hi to everyone's beautiful dogs and seeing the houses that look like bojack horseman's

if you think you're having a bad day, why, just look at the impropriety in which these youngsters have found themselves entangled

someone in the ravine area of prospect park is playing the recorder and i'm here for it (recordist not pictured bc i assume they are an invisible forest sylph)

(flattered to somehow be in this company?!)

the main trade-off of training a VAE is reconstruction fidelity vs. structure of the latent space, and the epoch 3 version of this model was much "better" at reconstructing arbitrary inputs. at epoch 12, it reconstructs inputs to appear more like the training data, which I kind of prefer? here's smashmouth again with reconstructions from this model https://gist.github.com/aparrish/d356dcfa7fddd57d4ffddf4f75f69399

picking a random point in the latent space, then doing greedy decoding from various randomly selected nearby points (essentially generating variations on a line)—

greedy decoding from random samples at epoch 12, just at the right point before the model started to collapse. these are actually sorta... breathtaking?

now training a variational autoencoder neural network on the gutenberg poetry corpus but with pre-trained subword embeddings—it's MUCH better at reconstructing inputs now—here's a reconstruction of a little number you might know

i don't know why people say software engineering is difficult. just get the code cube to unlock the area leading to the boss. even kirby understands the basics

me: [literally orders a single pair of shoes]

marketing email, running to catch up with me, desperate, breathing heavily:

hiked up to some pretty waterfalls on mount timpanogos today

COCO captions have to be the weirdest corpus commonly used in machine learning. just normal people describing normal things

computational poetics pedagogy 2k19 imo