Recently, I charted and started a new learning path toward mastery of machine learning. No sweat, right? Just put me in the chair and show me how to work the machine. I can paint houses, so… but no. It sounds crazy, but it turns out that if you want to understand what’s going on with machine learning, then some algebra and some statistics skills will only help. I’m okay with the algebra, but am currently enrolled in a class called Intro to Descriptive Statistics hosted by Udacity.com.
Meanwhile, TensorFlow is a totally free machine learning framework that Google has made available to the public, as well as a bunch of example uses and tutorials to get mere mortals into training a kind of artificial intelligence on their own problem domains. TensorFlow is apparently very good at learning how to recognize patterns in visual, textual, and audible data, as well as any known mathematical patterns in any data set. Machine learning comes in supervised training, unsupervised training, adversarial training – and every time you train up the AI in a new pattern-recognition skill, you can save that skill set, and add it like a single filter to a larger scheme of data processing.
The potential applications of this technology are out of scope. The next thing after this will be machines training other machines, so this looks like a good time to jump on board, if only to get an education in how to defeat the coming robot armies, or at least retrain them and send them into the past to protect me as a child. Personally, I’m more interested in how to train an integrated set of skills into a modern home equipped with an integrated set of sensors, including how to communicate easily and without errors, back and forth between home owner/occupant and home.
So, it’s the middle of 2017. By the end of 2017, I should have TensorFlow set up in the home and trained on a few toy examples. It would be great to also get a chance to do more analysis of sensor input data, and using recognized events to fire programming events, so fingers are crossed. My progress will be posted, as it happens.