View profile

MLT Newsletter - August 2022

Suzana Ilić
Suzana Ilić
Coming up: MLT news, resources and meetups. Share this newsletter with your friends for a machine learning upgrade! 🤖

Statistics lecture series: maximum likelihood estimation
These 4 lectures by Michal Fabinger cover a frequently used statistical method: maximum likelihood estimation. The concepts are introduced in an intuitive yet rigorous way. The topics include likelihood and conditional likelihood and their maximization, maximum likelihood estimation of linear models, logit models, models/logistic regression, probit models, and Poisson regression, score functions, Fisher information and asymptotic confidence intervals for model parameter estimates.
Michal Fabinger is the founder of the Acalonia school, which aims to build an education system for a world where location does not matter. Michal’s research is in physics and economics, with the corresponding PhD training completed at Stanford and Harvard. At the University of Tokyo and the Pennsylvania State University, Michal taught courses on Deep Learning, Data Science, Statistics, Asset Pricing, International Trade, International Finance, and Development Economics.
Past events
Text-to-Image Generation – Deep Dive into DALL-E 2, Imagen, Parti, VQGAN
Jerry Chi gives a general overview of the methods behind the recent trendy text-to-image models such as DALL-E 2 by OpenAI and Imagen and Parti by Google, and shows some fun resources for hands-on art generation.
Jerry Chi is a Staff Data Scientist working on job-jobseeker matching at Indeed Japan. Past jobs include analytics / ML at Google, Supercell, SmartNews, and his own failed startup. Taiwanese American; lived in Japan, China, Korea, Finland, etc.
AI at Scale in Rakuten’s Recommendation
Principal Data Science Engineer Phuc Le from Rakuten talks about how AI/ML is applied at scale in the e-commerce industry and gives a general introduction about recommendation systems. He dives into some modern types of recommendations used in Rakuten Group services and how multi-task learning with deep learning helps to improve buy-again recommendations.
Phuc Le is a Principal Data Science Engineer at AI Product Supervisory Department of Rakuten Group Inc. He received his PhD from the Tokyo Institute of Technology (Quantum Computation).
Rakuten is one of Japan’s top e-commerce companies providing 70+ different services with 1.6 billion members across the world.
A Practical Tutorial on Building Machine Learning Demos with Gradio
Building machine learning demos and web apps has traditionally required significant knowledge of web development (css, js) and web hosting. In this talk Abubakar Abid discusses the Gradio library, an alternative that allows you to build machine learning demos entirely in Python. This is a hands-on tutorial, be ready to code!
Abubakar Abid completed his PhD at Stanford in applied machine learning. During his PhD, he founded Gradio, an open-source Python library that has been used to build over 500,000 machine learning demos. Gradio was acquired by Hugging Face, which is where Abubakar now serves as a machine learning team lead.
👩🏻‍💻 Gradio Docs
If you haven’t subscribed yet to the MLT YouTube channel don’t forget to smash that subscribe button! 🧠
Interesting ML bits on Twitter
Stable Diffusion, Explained
How does Stable Diffusion work?
LLM.int8() is the first 8-bit inference method that saves 2x memory and does not degrade performance for 175B models by exploiting emergent properties.
Andrej Karpathy is back
with a 2h25m lecture on “micrograd”.
Hope this newsletter was useful! Until next time!
Did you enjoy this issue? Yes No
Suzana Ilić
Suzana Ilić @__MLT__

MLT Newsletter

In order to unsubscribe, click here.
If you were forwarded this newsletter and you like it, you can subscribe here.
Created with Revue by Twitter.