Roundup 11/1/2016

Machine learning (ML) and deep learning (DL) content from the past 24 hours. Plus, some AI stuff.

The Next Platform is publishing my three-part series on the state of enterprise machine learning. Part one is here. Part two is here.

IDC predicts worldwide spending on cognitive systems and AI will grow from $8 billion in 2016 to more than $47 billion in 2020, an annual growth rate of 55%.

Methods and Techniques

— In MIT News, Larry Hardesty summarizes an approach to making ML output more transparent and interpretable.

— Alex Castrounis offers a “complete and detailed” introduction to ML in five parts.

— Tony Fischetti explains the Bayesian approach to ridge regression.

— Syed Danish Ali and Rahul Ahuja write a long piece on handling structured and imbalanced datasets with deep learning. They apply DL with some success to the 1999 KDD Cup dataset.


— Google ports its image captioning models to TensorFlow, releases to open source.


— At NYU, Yann Lecun, Kyunghyung Cho, and Joan Bruna have a stupidly powerful NVIDIA DGX-1, and you don’t.




— Huawei uses ML to manage network traffic control.

— How to detect malicious URLs with ML.

— Haniya Rae profiles efforts by SK Planet, Amazon and Google to apply machine learning to shopping.

— Bob Violino surveys the use of ML against cellular network fraud as if it’s a new thing.

— NVIDIA wants to watch your baby.

Health and Medical

— A team in the UK uses wearables and ML to aid patient care for epilepsy.

— Israeli researchers use ML to predict resistance to antibiotics.

— A large medical practice in Boston uses ML on electronic health records (EHRs) to identify candidates for HIV pre-exposure prophylaxis.

— A team with advanced degrees attempts to predict hospital readmissions from two standard datasets with several ML techniques, discovers that they do not outperform logistic regression when measured by the C-statistic. Well, no shit, Sherlock! First, the C-statistic is explicitly designed for logistic regression, and it isn’t necessarily designed to compare the performance of different techniques. Second, successful efforts in rehospitalization modeling generally entail the inclusion of new data sources; one would not expect marked differences simply from using different techniques.


— Graphcore, a UK-based startup that develops accelerated chips for ML, lands $30 million in an “A” round. Linkapalooza here.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.