I was asked to give a guest lecture for a course, so I can pretend to be a real academic briefly!

ran out of oat milk so I made an affogato for my coffee of the day, that's just how life is when you're an adult

this project has the unfortunate property of being extremely cool and interesting, but also extremely not-my-main-area-of-research

Show thread

it turns out it's quite hard to *rigorously* say anything about the distribution arising from a messy stochastic process, especially when it's a side project and nobody involved is an expert on stochastic processes

Show thread

the project was intended to be a simple workshop paper in 2019 going "isn't this a cool idea", but we had the hubris to try to go beyond the "isn't this a cool idea" stage

Show thread

this week is dedicated to trying to "finish" my project on how stochastic gradient descent has randomness in it and maybe that provides privacy (in the sense of differential privacy)

Realising that I usually use conference travel as “holiday abroad” and that’s why I haven’t gone anywhere for so long

Embracing overleaf and becoming unimaginably powerful

I took Friday and Monday off and all I can say is: abolish Monday and also Friday

Virtual conferences have many upsides (accessibility, environment, cost) and a significant downside (actively draining to attend)

Show thread

it's H E R E
would have been cool if I could pick it up from Cambridge University Press in person but I guess that's not how supply chains work

Show thread

I need the opposite of an email amnesty day, where people remind me about their unanswered emails

ordered "Regression and Other Stories", can't wait to finally learn statistics

proud of myself for using dask to deal with a big file instead of just processing it line by like like I usually would. trying to get better at being strategically lazy with writing code

a more dramatic version of this was me being the last person to board a transatlantic flight becuase I *really* needed to finish setting my experiments running so they could finish before I landed

Show thread

my favourite thing about doing computational research is setting a long experiment running and then getting a coffee

I have some sense for how seemingly-innocuous questions asked repeatedly can drill into your bones, because "wow you don't sound Irish!" has been my life f o r e v e r

Show thread

back on my "trying to finetune deepspeech to handle my accent" bullshit.

had a moment where I thought fine-tuning it on a bit of Irish-language data would help, then remembered I barely sound like an Irish person


I am significantly less feverish and significantly more hungry for nuggets

Show thread
Show older
Scholar Social

Scholar Social is a microblogging platform for researchers, grad students, librarians, archivists, undergrads, academically inclined high schoolers, educators of all levels, journal editors, research assistants, professors, administrators—anyone involved in academia who is willing to engage with others respectfully.