View profile

nibble.ai weekly - Issue #21: 2019 review of the data/AI ecosystem, Retraining your entire org for AI & ML, fast.ai Deep Learning Foundations

Revue
 
Our top pick this week is Matt Turck's (FirstMark Capital) annual review of the data and AI ecosyste
 

nibble dispatch

July 3 · Issue #21 · View online
Curated essays about the future of Data Science. Production Data Science and learning resources for continuous learning. Covers Data Science, Data Engineering, MLOps & DataOps. Curated by people at https://nibble.ai/

Our top pick this week is Matt Turck’s (FirstMark Capital) annual review of the data and AI ecosystem. With the ecosystem evolving at lightspeed, this year’s is even split in 2 parts:
Definitely worth a read!
AI and Machine Learning will require retraining your entire organization
AI and Machine Learning will require retraining your entire organization
AI and ML will require retraining your entire organization
To successfully integrate AI and machine learning technologies, companies need to take a more holistic approach toward training their workforce. oreilly.com
AI and ML are going to impact and permeate most aspects of a company’s operations, products, and services. To succeed in implementing and incorporating AI and machine learning technologies, companies need to take a more holistic approach toward retraining their workforces.
News
Google releases the Youtube-8M Segments Dataset
This new extension to the famous Youtube-8M video dataset now includes human-verified labels at the 5-second segment level on a subset of YouTube-8M videos. googleblog.com
Albumentation v0.3
A new release for Albumentation, a fast image augmentation library. github.com
Wilcoxon Wars
In case you missed it, this went slighty viral on twitter
Very interesting twitter thread (+ analysis) about t-test vs Wilcoxon rank sum test against the perils of violated assumptions. einglasrotwein.github.io
Learning resources
New fast.ai course: part 2
While fast.ai part 1 was focusing on the practical aspects, this new course dives into the foundations (e.g. yes you should understand concepts like backprop) and uses Python, PyTorch and Swift for TensorFlow.
Here’s the release note and the course. fast.ai
The transformer … “explained”?
Attention is all you need
The transformer architecture is at the core of a lot of recent hot development in Deep Learning, especially NLP: GBT-2, BERT, and apparently AlphaStar aswell. A good first non-technical explanation to understand what lies behind the architecture. nostalgebraist.tumblr.com
The challenging part is in getting lots of good training data, and in finding a good training objective.
The “η-trick” or the effectiveness of reweighted least-squares
An article by Francis Bach, researcher at INRIA an alternative method to Newton’s method for convex optimization applied to non-convex functions. francisbach.com
Understanding the Bias-Variance Tradeoff
Oldie but goodie
Understanding how different sources of error lead to bias and variance helps to improve the data fitting process resulting in more accurate models. This article defines bias and variance in three ways: conceptually, graphically and mathematically.
For some reason, related to UTM parameters management, I can’t directly link to the page, here is the full link: http://scott.fortmann-roe.com/docs/BiasVariance.html
Did you enjoy this issue?
If you don't want these updates anymore, please unsubscribe here.
If you were forwarded this newsletter and you like it, you can subscribe here.
Powered by Revue