Machine learning and Deep Learning – what’s the difference?...Friday Morning Cyberamblings

Machine learning and Deep Learning – what’s the difference?

Machine learning -  simple definition:

Machine learning occurs anytime data is parsed by Algorithms (in order to learn from that data), and then apply that learning to better, more informed decision - making.

We know that parsing is a basic requirement to facilitate the needs of diverse entities who may need the data available in different formats. Parsing permits re-shaping data in a manner that can be interpreted by specific software, just like programs that are developed/written by humans, but ultimately, must be executed by computers.

How common has it become for you to browse the net in search of something you wish to purchase, only to find that the next time you visit your social media site, there are numerous, unsolicited recommendations for that self-same object of your desire? This is a great example of a machine learning algorithm developed around your browsing history and habits, which helps the online seller/service to make decisions about ‘target’ products they ought to recommend to you. Such machine learning algorithms may also associate your specific preferences with other browsers having similar tastes, to increase the breadth of their recommendations to you. This is a simple example of how AI is utilized in a plethora of services that offer automated recommendations.

Machine learning enables a variety of automated tasks that span multiple industries, from data security specialists that hunt/track malware, to finance professionals seeking early alerts on favorable trades. AI algorithms are programmed with the ultimate goal of continuous learning, and are capable of executing routine activities, while simulating the role of a virtual assistant —a job that they seem to perform rather well.

At the back-end, machine learning requires a lot of complex mathematics and coding that, in the final analysis, ends up serving as a mechanical function, much like a car, whose satellite navigation system takes coordinates you input, and then works out the best route to your destination, avoiding traffic snarl-ups along the way, and constantly looking for the quickest and shortest way to follow. Indeed, learning that when you say the words ‘directions home,’ it knows you want to go home and directs you accordingly.

Deep learning and machine learning compared

This topic gets more and more exciting when we learn about how machines can learn and develop new skills, through a combination of deep learning and neural networks.

In plain terms, deep learning is simply a subset of machine learning, and technically, incorporates similar functions. However, on closer examination we find that its capabilities are somewhat different.

Most simple machine learning models become gradually better at whatever they have been programmed to execute, but there may be instances where they may need direction. As an example, when an AI algorithm produces an inaccurate prediction, then a human being may be required to intervene and make appropriate adjustments. In contrast, with a deep learning model, algorithms are designed to ensure they are capable of independently determining the accuracy of their predictions, through their own neural networks.

If we go back to the car navigation system example: it could be programmed to provide directions home when it recognizes the audible cue of someone saying the words “directions home.” As it progressively continues learning, it might eventually provide directions home, with any phrases containing the word “home.” Now if car navigation system is endowed with a deep learning model, it could figure out that it should provide directions home with the cues “I want to go home” or, “take me home,” and in conjunction with satellite navigation, and the recognition of familiar route and street markers, it may be able to trigger directions home, when it senses that the route markers you are taking are similar to the route home. A deep learning system develops the capability of learning through its own, independent computing logic – almost like it can think and reason on its own – much like the human brain.

So how does deep learning function?

Typically, in deep learning models, a key design point is the facilitation of continual analysis of data, using a logic structure very similar to how we humans draw our conclusions. To achieve this objective, deep learning applications utilize a layered algorithm structure referred to as an artificial neural network. This artificial neural network is modeled after the biological neural network of the human brain, resulting in a learning process that is infinitely more capable (by order-of-magnitude), than that of standard machine learning models we are familiar with.

A great example of deep learning is IBM Watson versus the game “Jeopardy.” It triggered a wave of scientific and social consternation when IBM Watson defeated the game—not only could a machine master the complex nuances and many abstract factors of the game, but in fact it was good (through continuous learning and application) to beat Jeopardy – at its own game!