The StandardScaler is not Standard

  Рет қаралды 1,029

:probabl.

:probabl.

Күн бұрын

There is a scaler in scikit-learn called the "StandardScaler". The name might imply that it is standard or fairly basic., but once you consider the implementation details that are required for all the edge cases the it's actually far from "standard" . The goal of this video is to explain why. Hopefully, by the end of this video, you'll appreciate all the tiny details that scikit-learn handles under the hood for you.
00:00 Introduction
01:37 Documentation
07:03 Online Learning
10:43 Numerical Issues
14:11 Source Code
Documentation for the standard scaler can be found here:
scikit-learn.org/stable/modul...

Пікірлер: 3
@keanraw
@keanraw 9 күн бұрын
These deep dives are very useful! Thanks Vincent! I always tell my team to leverage existing tools, sometimes we like to think "oh, it's not that hard to implement X". And most of the time it actually isn't. However, we find ourselves getting to these edge cases fairly quickly, and then it becomes a whole thing. I'd love to see some scikit-lego related stuff.
@isbestlizard
@isbestlizard 16 күн бұрын
Oh my go dthat shade of blue and orange brings back memories of my data science course ahhhhh matplotlib in my dreams >.
@Sadjina
@Sadjina 16 күн бұрын
Now do random numbers drawn from a distribution with Pareto-tails with tail exponent < 2 ;)
The Quantile Trick
12:53
:probabl.
Рет қаралды 1,3 М.
Python Feature Scaling in SciKit-Learn (Normalization vs Standardization)
11:59
Barriga de grávida aconchegante? 🤔💡
00:10
Polar em português
Рет қаралды 45 МЛН
Who enjoyed seeing the solar eclipse
00:13
Zach King
Рет қаралды 137 МЛН
НЕОБЫЧНЫЙ ЛЕДЕНЕЦ
00:49
Sveta Sollar
Рет қаралды 7 МЛН
Why Does Diffusion Work Better than Auto-Regression?
20:18
Algorithmic Simplicity
Рет қаралды 6 М.
Pipelines for convenience, *and* safety
16:42
:probabl.
Рет қаралды 594
Probabl Livestream: Dipping Toes into Timeseries
1:00:35
:probabl.
Рет қаралды 928
NEW GPT-4o: My Mind is Blown.
6:28
Joshua Chang
Рет қаралды 330 М.
Probabl Livestream: Exploring Ibis for DataFrames
56:56
:probabl.
Рет қаралды 551
Probabl Livestream: KNN on a LanceDB Vector DB
47:00
:probabl.
Рет қаралды 302
How about that uh?😎 #sneakers #airpods
0:13
Side Sphere
Рет қаралды 9 МЛН
3D printed Nintendo Switch Game Carousel
0:14
Bambu Lab
Рет қаралды 1,7 МЛН
What % of charge do you have on phone?🔋
0:11
Diana Belitskay
Рет қаралды 312 М.