Dirty data science: machine learning on non-curated data

These slides are a one-hour course, touching on:
• missing values
• non-normalized categorical entrieshttps://t.co/zzlcCFcRMy

— Gael Varoquaux (@GaelVaroquaux) October 26, 2021

Image-Based CLIP-Guided Essence Transfer
abs: https://t.co/wFGwhJxRCZ
github: https://t.co/38xy9RrC9x

new method creates a blending operator that is optimized to be simultaneously additive in both latent spaces pic.twitter.com/WqKURD8ny8

— AK (@ak92501) October 26, 2021

Self-Supervised Learning by Estimating Twin Class Distributions
abs: https://t.co/LA6IagSCTv
github: https://t.co/QkBgV8FcRU pic.twitter.com/RW5OLHfb3W

— AK (@ak92501) October 25, 2021

A simple trick for learning a new literature I was taught in my undergrad is the "snowball" technique. Here's how it works: start with one review or empirical finding in the topic you want to work on. Carefully read the works cited and pick 4-5 key references. Then read those.

— Micah Allen (@micahgallen) October 25, 2021

SOFT: Softmax-free Transformer with Linear Complexity
abs: https://t.co/EralXVH5CZ
github: https://t.co/4miqmwAGcA

introduced a softmax-free self-attention mechanism for linearizing Transformer’s complexity in space and time pic.twitter.com/85Mw5MJOUc

— AK (@ak92501) October 25, 2021

The Arab Spring took place 10 years ago. This map shows how countries changed their ranking on the Democracy Index. Tunisia climbed up 90 ranks, Morocco 20, while Syria fell another 12. Source: https://t.co/1PG7VWZh7R pic.twitter.com/U7sTGHbIUZ

— Simon Kuestenmacher (@simongerman600) October 24, 2021

Police? I would like to report a dataviz murder pic.twitter.com/GkvcLO47d7

— Maarten van Smeden (@MaartenvSmeden) October 23, 2021

Here is a one-year perspective on @chrmanning's question (data courtesy of @SemanticScholar). Very interesting result: EMNLP has much fewer little-cited papers, but Findings has more very-highly-cited papers. Findings high-risk, sometimes high reward. 1/2 https://t.co/BnouEtU03e pic.twitter.com/d4Fj9Tv5YB

— Graham Neubig (@gneubig) October 21, 2021

Make sure you set PYTHONHASHSEED=0 *before* you start Python or Jupyter, as Python only reads it upon startup. Note that Google Colab sets it for you. pic.twitter.com/kjZMlj9pqz

— Aurélien Geron (@aureliengeron) October 21, 2021

FlexMatch: Boosting Semi-Supervised Learning with Curriculum Pseudo Labeling
abs: https://t.co/FDZTByplIY

FlexMatch outperforms FixMatch by 14.32% and 24.55% on CIFAR-100 and STL-10 datasets respectively, when there are only 4 labels per class pic.twitter.com/uMYQ171WoL

— AK (@ak92501) October 20, 2021

xFormers: Hackable and optimized Transformers building blocks, supporting a composable construction
github: https://t.co/HcAKAU2Mai pic.twitter.com/ZXTjcs70XK

— AK (@ak92501) October 20, 2021

How we think other people learn what they know versus how they really learn pic.twitter.com/Sme6JPtWJr

— Vicki Boykis (@vboykis) October 19, 2021