Why Deep Learning is Significantly Different From Machine Learning

Source: Deep Learning on Medium

Go to the profile of Rohit S

There is a great deal of confusion these days about Artificial Intelligence (AI), Machine Learning (ML) and Deep Learning (DL).

There certainly is an enormous uptick of short articles about Artificial Intelligence (AI) being a competitive video game changer which enterprises should begin to seriously check out the chances.

The difference between AI, ML and DL are extremely clear to professionals in these fields. Artificial Intelligence (AI) is the all including umbrella that covers everything from Good Old Fashion AI (GOFAI) all the way to constructionist architectures like Deep Learning.

ML is a sub-field of AI that covers anything that relates to the study of learning algorithms by training with data. There are whole swaths (not examples) of strategies that have been developed throughout the years like Linear Regression, K-means, Decision Trees, Random Forest, PCA, SVM

And, lastly Artificial Neural Networks (ANN). Artificial Neural Networks is where the field of Deep Learning had its genesis from.

Are you interest in learning “Deep Learning” you can enrol to best Coursera deep learning courses for more information

Some machine learning practitioners who have actually had previous exposure to Neural Networks (ANN), after all it was developed in the early 60’s, would have the first impression that Deep Learning is absolutely nothing more than ANN with numerous layers.

In addition, the success of Deep Learning is more due to the accessibility of more data and the schedule of more effective computational engines like Graphic Processing Units (GPU).

This naturally is true, the development of Deep Learning (DL) is basically due to these 2 advances, nevertheless the conclusion that DL is just a better algorithm than SVM or Choice Trees is akin to focusing just on the trees and not seeing the forest.

To coin Andreesen who said “Software application is eating the world”, “Deep Learning is eating Machine Learning (ML)”. Two of the publications by specialists of different machine learning (ML) fields have summarized it best regarding why Deep Learning (DL) is taking control of the world.

Christopher Manning a specialist in Natural Language Processing (NLP) writes about the “Deep Learning Tsunami

Deep Learning waves have lapped at the shores of computational linguistics for numerous years now, but 2015 appears like the year when the full force of the tsunami struck the significant Natural Language Processing (NLP) conferences.

Nevertheless, some pundits are anticipating that the last damage will be even worse.

Nicholas Paragios writes about the “Computer Vision Research Study: the Deep Depression

It may be simply because deep learning on highly complex, extremely determined in terms of degrees of freedom charts as soon as endowed with massive amount of annotated data and unthinkable — until extremely just recently — computing power can solve all computer system vision problems.

If this is the case, well it is merely a matter of time that industry (which seems to be already the case) takes control of, research study in computer vision ends up being a minimal academic objective and the field follows the course of computer graphics (in regards to activity and volume of academic research study).

These two short articles do highlight how the field of Deep Learning are essentially disruptive to standard Machine Learning (ML) practices.

Definitely it needs to be similarly disruptive in business world. I am however shocked and perplexed that even Gartner fails to recognize the difference between Machine Learning (ML) and Deep Learning (DL).

Here is Gartner’s August 2016 Buzz Cycle and Deep Learning isn’t even pointed out on the slide

What a travesty! It’s bordering on criminal that they their consumer’s have a myopic notion of ML and are going to be blind sided by Deep Learning.

Anyhow, regardless of being disregarded, Deep Learning (DL) continues to by hyped. The existing Deep Learning (DL) buzz tends to be that we have actually these commoditized machinery, that provided enough data and enough training time, has the ability to find out by itself.

This naturally either an exaggeration of what the advanced is capable of or an over simplification of the real practice of DL.

Deep Learning (DL) has over the past couple of years given rise to a huge collection of ideas and techniques that were formerly either unknown or known to be illogical.

Initially this collection of principles, appears to be fragmented and diverse. Nevertheless gradually patterns and methods begin to emerge and we are frantically trying to cover this area in “Style Patterns of Deep Learning”.

Deep Learning today surpasses simply multi-level perceptions however rather is a collection of methods and methods that are utilized to building composable differentiable architectures.

These are very capable machine learning systems that we are only today seeing simply the suggestion of the iceberg.

The crucial eliminate from this is that, Deep Learning may look like alchemy today, but we eventually will find out to practice it like chemistry.

That is, we would have a more solid structure so as to have the ability to build our learning machines with greater predictability of its abilities.