01:12:05
How to rethink the notebook - with Akshay Agrawal, co-creator of Marimo
09:00
Topics vs. embeddings
11:39
How the GapEncoder works
11:14
PCA as an embedding technique
12:08
Feature engineering for overlapping categories
01:09:11
You're always (always!) dealing with many (many!) tables - with Madelon Hulsebos
11:51
Data checks for estimators
11:22
Improving models via subsets
01:00:54
How Narwhals has many end users ... that never use it directly. - Marco Gorelli
10:13
Decayed estimators for timeseries
12:58
More flexible models via sample weights
12:47
Why ridge regression typically beats linear regression
12:15
Understanding how the KernelDensityEstimator works
01:05:39
Pragmatic data science checklists with Peter Bull
13:15
Use-cases for inverted PCA
10:48
Don't worry too much about missing data
01:01:48
Model safety, that's a pickle! with Adrin Jalali - scikit-learn maintainer
11:49
Boosting vs. semi-supervised learning
14:16
Benchmarking boosted trees against overfitting
12:43
Monotonic, and better, boosting
57:20
Moving towards KDearestNeighbors with Leland McInnes - creator of UMAP
10:24
Histograms for faster boosting
12:24
Getting deeper into trees
11:16
Why tree gradients give you a boost
01:04:10
Talk like a DataFrame, run like SQL with Phillip Cloud - core-committer on Ibis
16:32
The StandardScaler is not Standard
14:35
Scaling Datasets in Pipelines
01:11:56
Enhancing Jupyter with Widgets with Trevor Manz - creator of anywidget
01:31
Introducing: Sample Space
16:42
Pipelines for convenience, *and* safety
11:26
Building Elaborate Pipelines: Part 2
16:51
Building Elaborate Pipelines: Part 1
14:37
Random Search is better, but there is one caveat
16:55
GridSearch made faster with a Cache
08:32
Drawing a Dataset from inside Jupyter
13:42
Generating Periodic Features for Seasonal Timeseries
12:53
The Quantile Trick
09:42
Image Classification with scikit-learn
X
About
This is the official Probabl YouTube channel where we feature humans teaching and learning about machine learning, data science, and open source. More often than not, we will discuss scikit-learn, as well as a plethora of other tools and libraries to help data scientists, data engineers and data owners extract the most value out of their data.
Tags