Going Deep

What is deep learning?

Photo: Seeman. Source: Morguefile

Deep learning is a machine-learning method that computers use to recognize complex patterns, like human speech and 3-D objects. It takes a signal – like a sentence – and processes it through multiple layers of calculations and algorithms to come to a conclusion. It’s a way for a computer to “learn” something without being explicitly programmed. Of course, the computer has been programmed, just at a higher, more abstract level.

It uses probability to improve performance. A picture initially looks like it could be either a cat or a lion. But the picture comes from northern Europe, so the algorithm assumes it’s a cat and moves to the next level of evaluation. Deep learning algorithms are trained on a data set, then applied to new data. Remember when you could call Google to look up a phone number? Google used millions of phone calls – with feedback – to train their deep learning voice-recognition algorithm.

Deep learning is the discipline behind self-driving cars, natural language computer assistants like Alexa or Siri, computer-based music recommendations, and even medical diagnoses and drug discoveries. The algorithms use a broad array of correlations and massive databases to generate inferences and decisions, then use the feedback they get to refine their choices.

Deep learning algorithm. Source: Krizhetsky, Sutskiver, & Hinton

There are some things computers are really, really good at. They solve intractable problems that require massive data inputs with amazing speed, like a GPS app finding the fastest route through New York City traffic during rush hour, or directing freight trains across switching systems through Chicago. But there are some simple problems where they seem pathetically incompetent, installing a mandatory update just when I’m about to send out an email. Don’t they know I’m busy?!

It’s like designing a system to breathe underwater. An Aqualung is a complicated marvel of science and engineering. But breathing underwater is trivial … for a fish.

Douglas R. Tengdin, CFA

Charter Trust Company

“The Best Trust Company in New England”

By |2018-06-07T06:54:54+00:00June 7th, 2018|Global Market Update|0 Comments

About the Author:

Mr. Tengdin is the Chief Investment Officer at Charter Trust Company and author of “The Global Market Update”. The audio version of each post can be heard on radio stations throughout New England every weekday. Mr. Tengdin graduated from Dartmouth College, Magna Cum Laude. He received his Master of Arts from Trinity Divinity School, Magna Cum Laude and received his Chartered Financial Analyst (CFA) designation in 1992. Mr. Tengdin has been managing investment portfolios for over 30 years, working for Bank of Boston, State Street Global Advisors, Citibank – Tunisia, and Banknorth Group. Throughout his career, Mr. Tengdin has emphasized helping clients manage their financial risks in difficult environments where they can profit from investing in diverse assets in diverse settings. - Leave a comment if you have any questions—I read them all! - And Follow me on Twitter @GlobalMarketUpd

Leave A Comment