Machine Learning - Deep Learning

We will first take a look at a few deep learning applications to give you an idea of what it can do.


In several areas of machine learning, Deep Learning has shown great success.

Self-driving Cars − Autonomous self-driving cars use deep learning techniques to adapt to changing traffic conditions and improve over time.

The use of Deep Learning in speech recognition is another interesting application. There are a number of mobile apps that recognize our speech today. Apple's Siri, Amazon's Alexa, Microsoft's Cortena and Google's Assistant all use deep learning methods.

For organizing our photos, we use a variety of web-based and mobile apps. Face detection, face ID, face tagging, object recognition - all of these use deep learning techniques.

Untapped Opportunities of Deep Learning

People started exploring other domains where machine learning had not been applied so far after seeing the great success deep learning applications had achieved in many domains. Deep learning techniques have been successfully applied in several domains, and they can be used in many other domains as well.

  • People can use deep learning techniques to improve crop yields in agriculture.

  • Providing early detection of frauds and analyzing customers' ability to pay is another area where machine learning can be extremely helpful.

  • To create new drugs and provide personalized prescriptions to patients, deep learning techniques are also applied to the field of medicine.

As new ideas and developments emerge frequently, there are endless possibilities.

What is Required for Achieving More Using Deep Learning

To use deep learning, supercomputing power is a mandatory requirement. You need both memory as well as the CPU to develop deep learning models. Luckily, today we have an easy access to HPC – High Performance Computing. We can now see applications for deep learning that we discussed earlier being developed, and in the future we will be able to see applications in those untapped areas as well.

Let's examine some of the limitations of deep learning before we use it in our machine learning application.

Deep Learning Disadvantages

The following are some important points to consider before using deep learning.

  • Black Box approach

  • Duration of Development

  • Amount of Data

  • Computationally Expensive

In the next section, we will examine each of these limitations in more detail.

Black Box approach

In the following diagram, you see an application where you feed an animal image to a neural network, and it tells you that it is an image of a dog.

Machine Learning - Deep Learning

You don't know why the network came up with a certain result, which is why it's called a black-box approach. You're not sure how the network determined it was a dog. Think of a bank application where the bank wants to determine a customer's creditworthiness. You will certainly get an answer from the network to this question. Will you be able to justify it to a client? Why is the loan not sanctioned to customers?

Duration of Development

The diagram below shows how a neural network is trained

Machine Learning - Deep Learning

In order to solve a problem, you have to define it, create a specification, choose the inputs, design a network, deploy it, and test it. Take this feedback as a chance to restructure your network if it does not produce the expected results. This process is an iterative one and may take several iterations before the network is fully trained to produce the desired results.

Amount of Data

In contrast, traditional machine learning algorithms can be used with great success even with just a few thousand data points, as deep learning networks usually require a huge amount of data for training. The data abundance is growing at 40% a year and CPU processing power is growing at 20% a year, as shown in the diagram below.

Machine Learning - Deep Learning

Computationally Expensive

In order to train deep neural networks successfully, several weeks of training time may be required. Neural networks require several times more computing power than traditional algorithms.

The amount of computational power needed for training deep neural networks heavily depends on the size of your data and how complex and deep the network is, whereas traditional machine learning algorithms take only a few minutes/hours to train.

Here is a brief overview of Machine Learning, its capabilities, limitations, and applications.

Frequently Asked Questions

Ans: Machine Learning - Artificial Neural Networks view more..
Ans: Machine Learning - Unsupervised view more..
Ans: Machine Learning - Scikit-learn Algorithm view more..
Ans: Machine Learning - Deep Learning view more..
Ans: Machine Learning - Skills view more..
Ans: Machine Learning - Implementing view more..
Ans: Machine Learning - Conclusion view more..

Rating - NAN/5