Is there any need of Deep Learning?

This article was published as a part of the Data Science Blogathon.

Humans should be worried about the threat posed by artificial intelligence. – Bill Gates

I am sure that the above quote wants to convey some message to us, we should definitely think about it. What you think “Am I right?”, please share your opinion in the comment box, I will definitely read them which will help me to understand that “Is there any bad impact of these technologies on Human Species or not?” or “Will this technology be responsible for the extinction of the human species?”. 

Let’s begin, today our agenda is that we will discuss “Is there really any need for deep learning?”. 

In recent years we have probably been hearing a lot of hype about deep learning but what is it really all about & here is another curious question that comes to our mind that “Why has deep learning only now just come into the spotlight?”. Let’s first understand what actually Artificial Intelligence is.

Artificial Intelligence

Artificial Intelligence is an umbrella term for a branch of Computer Science. Its aim is for the machine to mimic human cognition, focusing on complex problem-solving. The only goal of AI is that the Machine will able to have human-like intelligence in the future. It refers to the simulation of human intelligence in machines that are programmed so that the machine will be capable of thinking like humans and mimic their actions.

Machine Learning is a subset of AI which basically focus on how to make a computer capable to learn on their own without the need of hand-coded instruction. Machine Learning systems analyze a vast amount of data and learn from previous mistakes. The results are generated from the algorithms that complete their task efficiently.

Deep Learning is a subset within Machine Learning, these technology attempts to mimic the activity of neurons in the human brain using matrix multiplication. This arrangement is called a Neural Network. Actually, the concept of Neural Nets comes into the picture in 1957 and was first tried in 1980 but it wasn’t proven useful. Deep Network is only become feasible because of only two reasons, 1st an increase in computing power and 2nd is a vast amount of data. After reading until this point, you will definitely doubt that whenever we are talking about Deep Learning every time “vast amount of data” this term comes with Deep Learning. Actually why Neural Nets need so much amount of data.

The answer to the above question is, actually more the data, the more robust your network will be. Due to robustness, your network will give more better and accurate results than any other algorithms. Let’s put ourselves in the shoes of Deep Learning😁. Suppose you have seen 3 pictures of the cat, taken from different angles. But on other hand, you have seen thousands of different cats, now you have a much easier time recognizing one. This is important for the data. In Deep Learning, data is the essence that allows the machine to learn.

Birth of Neural Network

The true field of Deep Learning began in 2012 bey before 2012, mos of the experts believed that Neural Network was useless. In 2012, Deep Learning explodes into the spotlight. In 2012, first-time Neural Network was used in the competition to recognize the world’s biggest image dataset and it actually blew all previous types of algorithms. At this movement, the world realizes the actual power of Neural Nets. This was the birth of the Neural Network.

My Opinion on “Is there any need for Deep Learning🤔?”

Before my opinion let’s look at a graph which is researched by google.

According to google searching after 2013 people mostly doing a lot of research on Deep Learning. The above graph is showing the interest of peoples in the field of Deep Learning. The graph has been increasing a lot after 2013 so what has led to this hike on-trend that we will understand in this article.

According to some experts, there are many reasons behind this trend or the exponential growth of interest of peoples in the field of Deep Learning. Let’s see them one by one.

1. First thing is that after 2013 most of the peoples being aware of the smartphones and start using smartphones. People start using various social media platforms such as Facebook, Instagram, or WhatsApp which actually generates a huge amount of data. By using this lot of data we can definitely do a lot of things, solve different kinds of use cases. Ex. recommendation system, and a lot of things.

Data is the main specific reason that Deep Learning comes into the picture. According to one survey each day approx. 2.5 quintillion bytes of data generated. Let’s look at one beautiful graph shared by Sir Andrew Ng.

Here above, we can see this specific graph where on the x-axis we have the amount of data and on the y-axis, we have the performance of the algorithm. As we see that as we increase the amount of data with respect to older learning algorithms(any kind of machine learning algorithms), the performance after a specific point of time, started degrading down and remains almost constant it did not increase. But in the case of Deep Learning, as we increase the amount of data, performance is also increasing. It means that this exponential growth of data led us to create some amazing deep learning models in terms of accuracy and various performance metrics.

2. Technology is another reason that encourages us to research Deep Learning because along with a huge amount of data, Deep Learning also required good quality hardware. Here I am talking about GPU(Graphics Processing Unit) and TPU’s(Tensor processing unit). Due to technology up-gradation, now we are easily getting good hardware at a very less price. As the technology is increasing day by day due to which the hardware cost is drastically decreasing day by day.

3. Actually Deep Learning is combining Feature Extraction and model training part. These two techniques we perform separately in the case of Machine Learning but here both the techniques are included within the deep learning techniques. Here feature extraction and model building which we do it separately in the case of Machine Learning is entirely combined in Deep Learning projects. Due to this Deep Learning can really solve complex problems such as image classification, object detection, or NLP task. Deep Learning actually uses the deep neural network, as the neural network becomes deep more and more complex information and features get extracted within a problem statement.

Let’s Wrap Up😅!

In the above points, we have discussed that Technology is continuously supporting us, so why not hold its hand and move one step ahead. But this technology has a set of significant disadvantages despite all its benefits. According to my, the Deep Learning model is incapable to provide arguments that why it reached a certain conclusion. I think it may cause some issues and it may be a challenge for Deep Learning Model. It’s ok that it requires a lot of time as well as good hardware to train the model.

I think that Deep Learning Models should also give a specific conclusion on its output, suppose whenever someone asks us that “is it a cat?”, the way we use to explain to him that why it is a cat, I think the Deep Learning Model should also come up with the same strategy that whenever it will give some output it will also give us a proper conclusion. I don’t think it is possible or not, I am only sharing my views😅.

And my final opinion is that actually, we should move ahead with trends and technologies, which actually help us to keep up to date and grow more. What you think, drop a comment below.

End Notes!

I hope you enjoyed this article. Any question? Have I missed something? Please reach me out on my LinkedIn. And finally, …it doesn’t go without saying,

Thanks for reading!

See ya!

The media shown in this article are not owned by Analytics Vidhya and is used at the Author’s discretion. 

Ronil