Thomas Bourgeois

Data science & Time series

[Thomas Bourgeois @__ThomasB]

@AntiTrumpCanada @HeyGenLabs French

[Thomas Bourgeois @__ThomasB]

@burkov https://t.co/QoUrn19fyu

[Yann LeCun @ylecun]

By releasing public demos that, as impressive & useful as they may be, have major flaws, established companies have less to gain & more to lose than cash-hungry startups.

If Google & Meta haven't released chatGPT-like things, it's not because they can't. It's because they won't.

[Thomas Bourgeois @__ThomasB]

@guolin_ke @GaelVaroquaux @scikit_learn *inspired

[Thomas Bourgeois @__ThomasB]

@GaelVaroquaux @scikit_learn I can do that

[Thomas Bourgeois @__ThomasB]

@PR0GRAMMERHUM0R An ex could still Ctrl-C this

[Thomas Bourgeois @__ThomasB]

@tng_konrad All I want for Christmas is autoregressive models explained by Konrad

[Thomas Bourgeois @__ThomasB]

@tng_konrad Original pic would be nice

[Thomas Bourgeois @__ThomasB]

@cory_eth @la_da_kid Who will have the guts to say out loud that chatGPT literally means in french : "Cat, I farted"

[Thomas Bourgeois @__ThomasB]

@GaelVaroquaux @deliprao It doesn't work on all algos as said in the tweet, but a colleague of mine says he got 10 to 30x on random forest

[Thomas Bourgeois @__ThomasB]

@Gorintic That's really amazing. Though you do slightly look like a wax statue out of Madame Tussaud.

Play opportunities seem endless with this kind of tool.

[Thomas Bourgeois @__ThomasB]

@svpino True. You shouldnt ever understand the details of that algorithm is another one.

[Thomas Bourgeois @__ThomasB]

@radekosmulski Unless I'm wrong this is classic Interview question : check of existence of a dict key/set element is in O(1) vs check of existence in a list is O(n).

Actually, used to not know this and did not get a job for this.. 😄

[Thomas Bourgeois @__ThomasB]

@jeremyphoward I'll be the first to use it when nbdev connects to Gitlab as well as to GitHub. I've wanted to use it for years but I'm stuck with my company's choice.

[Thomas Bourgeois @__ThomasB]

@CoinMarketCap df['users'].nunique()

[Thomas Bourgeois @__ThomasB]

@tng_konrad Data augmentation for TS can be useful (sometimes..)

https://t.co/iqa0WH19TV.

Rather than being general, I guess the
efficiency of such data augmentation can be evaluated, specifically per usecase (model + dataset), on a validation set.

[Thomas Bourgeois @__ThomasB]

@PierreDeWulf Though that seems very "sensé"

[Thomas Bourgeois @__ThomasB]

@PierreDeWulf Kind of tweet that can get you in lots of trouble ;)