Some common misconceptions about deep learning
It is not uncommon to find misconceptions in many areas of science & technology, for example, big bang, which is considered the origin of the universe is one of them. In the fields of Artificial Intelligence (AI) and Machine Learning (ML) I think “deep learning” will be at the top when it comes to misconceptions.
It is not that only beginners have misconceptions about deep learning, many top companies have misleading content on their web pages about deep learning. One of the obvious reasons is that this misleading content was not written by the experts or practitioners but marketing people. Here in this article, I will give a list of misconceptions which I commonly come across.
1. Deep learning & machine learning are two different branches of artificial intelligence and some of the problems can be solved only either by deep learning or machine learning. This is really surprising because the diagram which shows machine learning as a sub domain of artificial intelligence and deep learning as a sub domain is very well advertised.
2. Deep learning does not always involve neural network. This is another major misconception where people keep discussing about deep learning without mentioning artificial neural networks which make the backbone of deep learning. Without neural networks there is no deep learning.
3. Deep learning is much more complex than machine learning. This misconception is also very common but the fact is that many of the machine learning algorithms, such as support vectors machines or SVM, are much more complex than deep learning.
4. Machine learning needs a lot of data however, deep learning can work with a small amount of data. This is strange because the fact is that deep learning models need much more data than other machine learning algorithms, mainly because they discover features themselves.
5. Deep learning is about deep understanding of the data! Good number of candidates I interview fail to answer what is “deep” in deep learning. In deep learning deep is associated with the number of stacked layers and not any deep meaning.
6. Deep learning needs a very special kind of skills & mathematics. This is completely wrong, deep learning needs exactly the same linear algebra & calculus which other machine learning algorithms use.
7. Deep learning can solve any problem. Again, this is a common conception because there are many problems for which other machine learning algorithms may be much more useful than deep learning.
8. Deep learning is the most advanced form of machine learning which was discovered in the recent years only.
After discussing the misconceptions let me mention some facts &figures about deep learning. Let us discuss some of the official definitions of deep learning.
Deep learning is a subset of machine learning, which is essentially a neural network with three or more layers. These neural networks attempt to simulate the behavior of the human brain — albeit far from matching its ability — allowing it to “learn” from large amounts of data. While a neural network with a single layer can still make approximate predictions, additional hidden layers can help to optimize and refine for accuracy [IBM]
Deep learning (also known as deep structured learning) is part of a broader family of machine learning methods based on artificial neural networks with representation learning [Wikipedia]
Deep learning is a form of machine learning that enables computers to learn from experience and understand the world in terms of a hierarchy of concepts. Because the computer gathers knowledge from experience, there is no need for a human computer operator to formally specify all the knowledge that the computer needs. The hierarchy of concepts allows the computer to learn complicated concepts by building them out of simpler ones; a graph of these hierarchies would be many layers deep [Deep learning book by By Ian Goodfellow, Yoshua Bengio and Aaron Courville]
Deep learning is a specific subfield of machine learning: a new take on learning representations from data that puts an emphasis on learning successive layers of increasingly meaningful representations. The deep in deep learning isn’t a reference to any kind of deeper understanding achieved by the approach; rather, it stands for this idea of successive layers of representations. How many layers contribute to a model of the data is called the depth of the model. Other appropriate names for the field could have been layered representations learning and hierarchical representations learning. Modern deep learning often involves tens or even hundreds of successive layers of representations — and they’re all learned automatically from exposure to training data. Meanwhile, other approaches to machine learning tend to focus on learning only one or two layers of representations of the data; hence, they’re sometimes called shallow learning [Deep learning with Python by Francois Chollet]
Deep learning is a subset of machine learning that’s based on artificial neural networks. The learning process is deep because the structure of artificial neural networks consists of multiple input, output, and hidden layers. Each layer contains units that transform the input data into information that the next layer can use for a certain predictive task. Thanks to this structure, a machine can learn through its own data processing [Microsoft] .
Most of the definitions mentioned above make it very clear that deep learning is based on artificial neural network and the backbone of which is back propagation algorithm. Let me end this article by giving three of the most important features of the deep learning.
- 1. Deep learning models learn the data representation in terms of featured automatically.
- 2. Deep learning which employs layers of artificial neurons can be used for classification as well as regression.
- 3. Since deep learning models involved many layers of neurons so it is quite common to have millions of fitting parameters (weights) and that means we need a lot of (labelled) data.
- 4. Since deep learning needs a lot of data to be processed so we need very high performing computing systems (such as GPU/TPU) to training any deep learning pipelines.
In case you find this article please like, comment & share.
