Patrick C

Data scientist, machine learning enthusiast

[Patrick C @data_pat]

@alex_gude Yeah although I wish I had something even more ephemeral for that tbh, the temptation to go back and "reuse" the saved notebook leads down a dark path

[Patrick C @data_pat]

thinking of reviving this account, as a "notebooks are almost always the wrong tool" rageposting venue

[Patrick C @data_pat]

@ymt1234 Horror movie pitches for data scientists

[Patrick C @data_pat]

@JoFrhwld Whatever you do make sure to cover the philosophy of social science too since there are some interesting conversations there, which are relevant to lx dep who you're convinced by. Habermas On the logic of the social sciences, Foucault The Order of Things, but idk any great inros

[Patrick C @data_pat]

@adotheo You're not my real dad

[Patrick C @data_pat]

where do i shout about people using "deprecate" to mean "decommission"? "We're deprecating this service in July" No you're deprecating it right now-- that's exactly what you're doing when you say you're going to turn it off~~~

[Jacob Eisenstein @jacobeisenstein]

Trained up a chatbot on 100M turns of reddit comments and replies from 2010. I'm pretty sure I won the Turing test, who do I call to collect the prize?

[Patrick C @data_pat]

Wherein I compare a classic feature selection algorithm to our homegrown alternative and find *strengths and weaknesses for both*!! Stunning~ #datascience #machinelearning https://t.co/goN5AMY1I5

[Patrick C @data_pat]

No shade to the jupyter lab devs but really hoping some interface ~softening~ is on the roadmap. Like the new layout but the menus to run cells and restart kernels etc are so crowded!

[Patrick C @data_pat]

@data_pat and thanks to surprisingly consistent hype person @alex_gude for reposting this every once in awhile... check's in the mail

[Patrick C @data_pat]

"Sadly, in the real world hello is closer to goodbye than it is to a hamburger" --my masterwork https://t.co/p3PDCaqOFd

[Patrick C @data_pat]

You know @alex_gude, years into my python career, I keep getting burned by the *implicit* None returned by functions that I've clearly ended with the expression that should carry the function's value~ #rstats #forever

[Randy Olson @randal_olson]

It's official: Starting January 1, 2019, pandas will drop support for #Python 2. This includes no backports of security or bug fixes. #programming #DataScience

https://t.co/QeEC0hLjq4

[Patrick C @data_pat]

@E_J_White Verifies that it can learn anything at all. My preprocessing is at least adequate, I didn't forget the last nonlinearity, most of the right pipes are all hooked up.

[Patrick C @data_pat]

Nothing like that warm fuzzy feeling of finally getting it to overfit (to debug the process) before settling into the dread of actually developing a model that performs #machinelearning #gru #deeplearning

[Patrick C @data_pat]

if you don't vote you can't complain
/ɯ/ /ɯ/ /ɯ/ https://t.co/Xg6wBkRZ7B

[Tyler Schnoebelen @TSchnoebelen]

The choice here is obvious. Yes, I know some of you like your tongues forward, but come on, it’s way more fun if you make an ooo sound but ***unround*** your lips. Vote for whimsy! Vote for /ɯ/! https://t.co/qIoGZfHLoG

[Patrick C @data_pat]

you thought GANs were a toy, but using them to mock up complicated probability distributions can help answer some tough what-if questions on real world data! https://t.co/26XHwovJXz

[Patrick C @data_pat]

@TSchnoebelen 'unpredictably moving' makes kangaroos sound like the next indie sleeper hit

[Patrick C @data_pat]

@data_beth good to know!