Machine Learning on Steroids

As the Science and Technology Advisor for Applied Technology Institute, I am assigned the task of writing a weekly blog to discuss some upcoming course being offered by ATI.  I am supposed to be knowledgeable in most areas, so imagine my surprise when I was directed to write about our upcoming course entitled Deep Learning […]

As the Science and Technology Advisor for Applied Technology Institute, I am assigned the task of writing a weekly blog to discuss some upcoming course being offered by ATI.  I am supposed to be knowledgeable in most areas, so imagine my surprise when I was directed to write about our upcoming course entitled Deep Learning Architectures for Defense & Security.  Due to the nature of my background, I was unfamiliar with this topic, but I did know it was somehow related to Machine Learning and Neural Nets.  I dutifully went to old textbooks and google.com to learn as much as I could.  I really wanted to find a single, concise, explanation which would make me instantly smart, but no such single, concise answer was to be found.  So, for anyone else who needs that concise explanation, let me tell you what concise explanation would have been helpful to me before I spent hours on this topic.  Machine Learning is the most basic form of Artificial Intelligence, and Deep Learning is Machine Learning on steroids.  It’s just that simple, but let me explain. 

Machine Learning is a form of artificial intelligence which allows the machine to train with diverse datasets, and then predict based on those experiences.  Machine Learning uses automated algorithms to predict decisions for the future.  Most importantly, all of the analysis and algorithm tweaking is supervised by a human being whose goal is to improve the quality of the Machine’s performance.  One Example of Machine Learning would be speech Recognition, where an algorithm hears a spoken word and determines what that word is.  Through Machine Learning, the Engineer will help the algorithm improve, and get more and more words correct over time.

With Deep Learning, there is a constant focus on improvement and flexibility by using Algorithms which duplicate the thinking of the human brain, and do not require the constant supervision of a human.  Through use of multiple layers of neural networks ( yep, neural, just like in the human brain ), the algorithms learn to make decisions as a human would, or perhaps better than a human would.  To go back to our example from above, while Machine Learning would be used for an application like voice recognition, Deep Learning and neural nets would be used to go much further, such as with virtual assistants like Alexa or Cirri.  The algorithms must not only covert the spoken word, but the algorithm must understand the meaning of questions, and then compose an intelligent response.  Deep Learning is Machine Learning on Steroids.   

As daily life becomes more complicated, as systems become more complex, and as manpower becomes harder to secure, the importance of Deep Learning will continue to rise.  Today’s engineers need to continue making advances in Neural Nets and Deep Learning, so that “toys” as well as our essential machinery will be able to support the future needs of mankind. 

If you would like to learn more about how Deep Learning can be applied to your work in Defense and Security, please consider taking the ATI short course entitled Deep Learning Architectures for Defense & Security.  You can learn more about this course, and register for it, here.

Also, you may want to take a look at the full list of upcoming courses being offered by ATI here.