Reviews on AW — notes & thoughts

Sparse notes

Wed, 01 Mar 2023 21:28:43 -0500

Data science

march Effect of irrelevant features - a nice experimental test
march Matrix factorizaton recap in 6minutes
march mean encodings: a comprehesive study, check also category_encoders sklearn library
march relationship between SVD and PCA + a comprehensive tutorial on arxiv
march Naive-Bayes classification from scratch
On Embeddings for Numerical Features in Tabular Deep Learning
Euclidean vs cosine distance

Data viz

Complimentary

Temporary

Staging area for links that will be forgotten soon:

march H&M personalized fashion reccomendation
march learning D3js + essential training + custom visuals into powerbi

Q1 2023: Miscellanea (playlist)

Wed, 01 Feb 2023 21:28:43 -0500

Usare i numeri ciecamente: la fallacia di McNamara

Due note brevi sulla fallacia di McNamara (segretario della difesa USA durante la guerra in Vietnam), spesso riassunta come “fallacia quantitativa”

“the facts which count best count most". [I fatti che si contano meglio sono quelli che contano di più]

Spesso è figlia delle organizzazioni, ma colpisce (duro) individui e decisioni. Come fare a riprodurla? Ecco un tutorial (preso da qui):

As reported by Smith, Yankelovich describes the fallacy as follows:

The economics of the agri-food value chain

Tue, 01 Mar 2022 21:28:43 -0500

Course review

Let me paraphrase what Tukey once said: I truly believe that getting in contact with different fields and expertise is one of the perks of being a data scientist. Pictorially:

The best thing about being a ~~statistician~~ data scientist, Mr Tukey once told a colleague, is that you get to play in everyone’s backyard.

So, even if you can legitimately play in everyone’s backyard - which oftentimes means cleaning it as a first step, it also means that you will be speaking to people with different backgrounds ad expertise.