The definitive weekly newsletter on A.I. and Deep Learning, published by Waikit Lau and Arthur Chan. Our background spans MIT, CMU, Bessemer Venture Partners, Nuance, BBN, etc. Every week, we curate and analyze the most relevant and impactful developments in A.I.
We also run Facebook’s most active A.I. group with 191,000+ members and host a weekly “office hour” on YouTube.
Editorial
Thoughts From Your Humble Curators
This week we cover two new open source frameworks: the fast.ai Pytorch library and Microsoft infer.NET.
As always, if you find this newsletter useful, feel free to share it with your friends/colleagues.
This newsletter is published by Waikit Lau and Arthur Chan. We also run Facebook’s most active A.I. group with 176,000+ members and host an occasional “office hour” on YouTube. To help defray our publishing costs, you may donate via link. Or you can donate by sending Eth to this address: 0xEB44F762c58Da2200957b5cc2C04473F609eAA65.
Join our community for real-time discussions here – Expertify
News
Project Maven: Redux
There are plenty of image classification applications in the military. So it should not come as a surprise that Google had been working with them. Google walked away from Maven back in June, would they come back?
Blog Posts
fastai for PyTorch
We have great respect for Jeremy Howard who developed the very popular and practical course, aptly called Practical Deep Learning for Coders. And now, Jeremy is also publishing the course’s library which build on top of Pytorch. It has optimizations and blessings from the original Pytorch team, and has already been showcasing itself in several commercial projects. This post from Jeremy highlights several recents projects using the library, and what’s under the hood.
PyImageConf 2018 Recap by Joey Rosebrock
This is summary of PyImageConf 2018, written by one of the hosts: one of our favorites in our community, Joey Rosebrock. While this is a new conference, Joey was able to invite several luminaries in deep learning, such as Francois Chollet and David King to the venue.
Kudos to Joey and his co-host Jeff Nova. Keep up the good work!
MIT Human-Centered Autonomous Vehicle
Here is the paper version of human-centered autonomous vehicles we mentioned last week.
Open Source
Microsoft Infer.NET
Our first impression of Infer.NET was “Oh, isn’t this yet another dot net product?” And “MS already has CNTK” already. So why one more framework? But we were pleasantly surprised. Infer.NET is more about what the post termed “model-based machine learning” which is really about training up distributions such as good old classics like mixture of Gaussian distributions, principal component analysis. So in a sense, it’s more a “machine learning framework” rather than “deep learning framework” as we now accustomed to. You won’t see the now standard tutorial on DNN, image recognition and machine translation, but then you will find interesting applications such as recommendation systems or skill assessment.
Infer.NET was used in many real life cool applications, e.g. as cited by the post, is the Trueskill 2 system which access capability of players. There was also hundred plus papers based on the work.
This is already quite refreshing. The team wrote a book about the work in their book “Model-based Machine Learning” and it is cowritten by one of our favorite authors, Christorpher Bishop who penned the Bible “Pattern Recognition Machine Learning”. Furthermore, there is a future chapter on probabilistic programming. Wowowow, there are just too many goodies here. So don’t forget to check it out.
About Us
Join our community for real-time discussions here: Expertify