Profile cover photo
Profile photo
Mark Watson
817 followers -
I am an author and consultant specializing in artificial intelligence, semantic web, Ruby, Lisp, and Java
I am an author and consultant specializing in artificial intelligence, semantic web, Ruby, Lisp, and Java

817 followers
About
Communities and Collections
View all
Posts

Post has attachment
Very cool: using the library IntegratedGradients to explain predictions made by deep learning classification models. I used the version for Keras to update one of my old deep learning examples today: http://blog.markwatson.com/2018/02/trying-out-integrated-variants-library.html
Add a comment...

Post has attachment
My simple way to schedule machine learning jobs to run in sequence: http://blog.markwatson.com/2017/12/scheduling-machine-learning-jobs-to-run.html I do this to schedule Keras (or TensorFlow) jobs. This might be useful to you and save you a few minutes of research.
Add a comment...

Post has attachment

Post has attachment
As someone who is a generalist, with a few deep specialties, I loved this article: https://www.farnamstreetblog.com/2017/11/generalized-specialist/

I manage a machine learning team and more than once I have recommended also having deep knowledge of different aspects of software engineering that hit individual interests.
Add a comment...

Post has attachment
My notes for easy Jupyter notebook setup on AWS GPU EC2 with machine learning AMI: http://blog.markwatson.com/2017/11/easy-jupyter-notebook-setup-on-aws-gpu.html Amazon did a good job setting up GPU support and just about all major deep learning frameworks on their machine learning AMI

I considered getting a Linux box with a GPU for use at home but given my light requirements (less that 10 hours a month on my personal projects), it is much less expensive to rent by the hour.
Add a comment...

Post has attachment

Post has attachment
catching up: http://blog.markwatson.com/2017/09/new-job-and-two-deep-dives-into-tech.html

I had not written a blog in 4 months, being busy with moving and a new job.
Add a comment...

Post has attachment
Apple's new machine learning SDK for iOS looks promising https://developer.apple.com/machine-learning/ - perhaps a topic for my next book?

I wrote an iOS app prototype for nutrition this year using RubyMotion but I would like to try Swift and XCode. Apple's machine learning APIs, pre-trained models, and tools for training models all look good and offer me some incentive for climbing up the Swift and XCode learning curve.

I have 30 years (not full time) of working experience in AI and machine learning and I am always on the lookout for applications.
Add a comment...

Post has attachment
I updated my NLP library for Pharo Smalltalk: http://blog.markwatson.com/2017/05/i-updated-my-natural-language.html

I first used Smalltalk in 1983 on my Xerox Lisp Machine. Fun times.
Add a comment...

First time with Jupiter notebooks/kernels - Common Lisp cl-jupyter (link: https://github.com/fredokun/cl-jupyter) github.com/fredokun/cl-ju…) so far is nice to setup and easy to use. After watching a video of Peter Norvig's recent talk in NYC about the future of development tools (both programming and data science) I decided to do a one month trial of Wolfram Language. While Wolfram Language is certainly very nice, I started thinking about how I might create a very tiny subset with just the things I want/need. So far, Jupyter with a Common Lisp kernel is the most promising platform I have found to build on.
Add a comment...
Wait while more posts are being loaded