Have You Heard About Deep Learning Before - To You About It ...
This article is intended to address completely new people in deep learning who are planning to enter this field. The purpose of this article is to help them think and familiarize them with its complexity and help them distinguish between trivial and difficult things.
The general advice that should be given in this article is that deep learning is very easy, so choose something more difficult to learn because learning neural networks should not be the goal but rather it should be anything side.
Deep learning is a powerful science because it makes difficult things easy
Deep learning allows us to formulate many of the problems that were previously impossible into a conceptually simple matter, as deep networks deal with natural signs that we did not have easy ways to deal with in the past such as images, video, human languages, speech and voice where everything we do with the help of learning Deep eventually becomes very simple without any excessive thinking about the way to solve, besides there are high level platforms ready to help like TensorFlow, Theano, Lasagne, Blocks and others.
Choose something more difficult than deep learning
Deep Learning uses innovative methods, as Generative Adversarial and Variation Autocoders are great examples that have aroused great interest in probabilistic modeling, an understanding of how they work and construction around understanding the cause and how these deep neural networks work.
Building deep neural networks for supervised learning is now boring or solved by many, and therefore many of them have gone towards learning without supervision and making use of a set of new tools, besides you must have a very different kind of thinking and familiarity with information theory / probability and engineering .
Deep learning currently means that most people use relatively simple tools, but within six months many of them will acquire a set of skills, so do not spend much time working on getting to know things that will seem over time very easy, as you will miss your opportunity to make a real impact in Your job and distinguishing your career in the long run, just think about what you really want to learn and choose something else more difficult then go on to work with people who can help you with that.
Refer to the basics
Learn things like EM algorithm, Variation interface, non-supervised learning with linear Gaussian systems like PCA, factor analysis, Kalman filtring, and slow feature analysis.
While deep learning has become interesting recently, it is useful to try to bet on other areas that will gain importance in the future, such as:
Probability programming and probabilistic induction of a black box with or without neural networks. MCMC and variable inference methods with or without deep neural networks are also used.
Image credit
Have you seen this before?
The same thing happened several years ago when they used Hadoop and Hive .... And so on, as their use was important, and many of their first users achieved brilliant and wide success, all you have to do is classify things into small groups distributed only only ... and soon you will find tens of thousands of followers of you who will sanctify you simply because they understand that you are a leader in a world data.
It looked magical at the time, but looking back a few years ago you'll find it trivial, there are a lot of people using Hadoop and Spark now and tools like Amazon Redshift that have made things even simpler, it's all about how to use these tools, in case you want Hadoop is used by your company but for an error unless this tool is used correctly, this technology feature will evaporate very quickly.
Because there are training courses in data science, online training programs, training camps, etc., many people have graduated from these programs until these skills have become somewhat trivial, and what is happening now with deep learning looks quite similar.
In short, if you are just about to engage in deep learning, think about what that means, and try to be more specific with what you want, think about how many people you have now put them in your position, and how you will be sure of the things you will learn and not those that seem boring or trivial within a year.