The Delta-V Awards: Machine Learning

machine learning awardsThe ESG Delta-V awards recognize the top 20 companies that made an impact in big data and analytics in 2015. Here are the ones to congratulate for their success in the category of Machine Learning, including cognitive, artificial intelligence, and variants.

Learn a bit about the Delta-V awards here.

IBM Watson — The biggest problem with Watson is the reflexive skepticism that this can't be a real product. To be fair, it's non-trivial to get Watson delivered and trained up, but as a solution, this is the closest it gets to sci-fi starship-board computers like HAL 9000 and the Heart of Gold. Except instead of barring airlocks or making tea during missile attacks, Watson is diagnosing medical problems, powering business applications, and providing associative knowledge on demand. One of the best features is the collaboration between man and machine: if you doubt an answer, check the confidence rating and offer other suggestions. It learns from human expertise as much as data analytics.

Google TensorFlow — Maybe a bit controversial, initial reactions to TensorFlow have been mixed, but there are two unarguable points here.

  1. Google knows how to do machine learning on vast amounts of data.
  2. Google is now sharing that knowledge with the world.

You can have free access to libraries and tool sets that offer great flexibility for development, environments, and programming languages. Performance is scalable and tunable, and it's easy enough to see what matters. While early days yet, this should join other open source machine learning options (like Spark MLlibd, Apache Mahout, etc.) and with Google's guidance get more people on board with ML. 

Skytree — Here's the thing about Skytree: your genius machine learning PhD's could write the custom machine learning logic using open source tools, and given a couple of days most likely do it pretty well. If you had any genius machine learning PhD's. Also assuming you'd want them to spend a couple of days on the model instead of using Skytree and getting the same results in 20 minutes. Of course, you don't just model once either, you do this iteratively, so keep multiplying that 2 days vs. 20 minutes by the number of times you want to use machine learning in your enterprise. Helpful hint: just use Skytree already. You don't lose any control, accuracy, or specificity, you just get there way sooner.

devops info

Topics: IBM Big Data google machine learning