Uhm... if I get a decision tree like the one shown in the picture, does it mean that I only need the columns shown in the tree for training and validation, right? I would only need the columns 2 and 3 (x[2], x[3]), isn't it? Or am I missing something else?

When training a model it turns out that I get better results with a small dataset than with a bigger dataset. This is what is called overfiting, right?
#MachineLearning #Sklearn

Dear Machine Learning people: when a problem can be solved using both a regressor and a classifier, which method would you choose? Or you simply try both and then choose whatever worked better? Any rule or set of rules to try to determine which method should work better?

I ran a quick Gradient Boosted Trees vs Neural Nets check using scikit-learn's dev branch which makes it more convenient to work with tabular datasets with mixed numerical and categorical features data (e.g. the Adult Census dataset).

Let's start with the GBRT model. It's now possible to reproduce the SOTA number of this dataset in a few lines of code 2 s (CV included) on my laptop.

1/n

Lots of talks on the topic of AutoML. The #KDD2023 Test of Time Award Presentation by Frank Hutter gave a nice overview of the evolution of the #AutoML field. #AI #weka #sklearn #AutoWEKA #AutoSklearn #datamining

Somebody wrote here that "𝘮𝘶𝘴𝘪𝘤 𝘪𝘴 𝘫𝘶𝘴𝘵 #𝘮𝘢𝘵𝘩𝘦𝘮𝘢𝘵𝘪𝘤𝘴, 𝘣𝘶𝘵 𝘭𝘰𝘶𝘥𝘦𝘳". #Math is indeed beautiful.

Analyzing high-dimensional data (10^6 samples, 100+ features) in #Python, using #Sklearn Looking at that data thru a manifold, reducing the number of dimensions to 2 (e.g. a sheet of paper), sometimes to 3 (#3D model), with the help of #TSNE method.

It doesn't cease to amaze how many living creatures, leaves, insects I see in this unreal world

Links:

https://en.wikipedia.org/wiki/Manifold

https://www.jmlr.org/papers/volume9/vandermaaten08a/vandermaaten08a.pdf

What’s the best way to curate your feed on Mastadon? 
My interests include: #machinelearning #datascience #talmud #graphtheory #nlp #deeplearning #GraphNeuralNetwork
Languages: #python #r #rstats #golang #rust #cypher
Projects: #pandas #polars #numpy #sklearn #neo4j

I just did an #introduction a few days ago, but I've moved servers, so let's try one more time, for the cheap seats in the back!

I'm currently a data analyst/product #DataScientist working with free-to-play #VideoGames, and living in #Halifax, #NovaScotia, #Canada. I've done a lot of work on #Analytics design, with a focus on ensuring player telemetry events are sensibly cross-referenceable, and looking for relationships between engagement with different game features and business outcomes.

Business teams in freemium games love looking for magic buttons.

I primarily use #SQL, #Pandas, #Statsmodels, and #SKLearn on #Databricks (#Python), and #JuliaLang (DataFrames.jl, GLM.jl, Gadfly.jl, and Makie.jl, etc) for smaller, locally run projects. My interests lie in expanding the library of ML models I have in my back pocket for performing inference based knowledge generation. I'm not super keen on automating products with quasi-black-boxes for the sake of revenue optimization. If I'm not personally learning something new about people through my work, I don't usually see the value in it.

I did my BSc in #Physics and my MSc in #Astronomy, and, though I had dreams of progressing further down that pipeline, life kind of got in the way. Between the two degrees, I worked at the #Edmonton #Planetarium for four years as a presenter/operator (should out to the #ZeidlerDome at #TWoSE!).

I'm a life-long #Trekie, thanks to my mother. I grew up with #TNG and #DS9, and watched the first 5 seasons of #VOY before leaving home for university. Currently very bullish on #SNW and #LDS.

I'm also a lifelong #Baseball fan (#BlueJays and #Expos), and actively play rec #Softball.

About a year ago, I purchased my first lens-swappable digital #camera, and have been figuring out #Photography ever since. Most of my posts have focused on sharing my pictures, though I've recently decided to start a dedicated account for that on a #PixelFed server, for the sake of searchability.

My wife is currently studying political sociology, and I find her work fascinating. She's not currently on the Fediverse, but maybe one of these days.

This has really lost all sense of narrative flow, hasn't it? Oops!