The term "Machine Learning" dates as back as year 1959 when Arthur Samuel then working at IBM coined it. Nearly six decades after that, due to advancements in computing and optimizations in number crunching, we could see it practically happening. But what exactly does the Machine Learning brings us? Well it brings us the "Artificial Intelligence", the process of mimicking the way humans learn, perceive and make decisions. It employs various methodology of gathering inputs, analyzing data, recognizing patterns and making strategies. But then what is Deep Learning and how, if at all, it is different to Machine Learning?
Machine Learning (ML)
Machine Learning involves development of a system that can "learn" itself when provided with input just as humans do. The system employs algorithms to analyze the incoming data stream and to identify structures and patterns. These algorithms evolve and becomes more efficient over time when presented with diversified and large amount of data.
What do you think is written in the image on the left? Many people instantly recognize it as number 435679. This is the wonder that our human brain does which is evolved over thousands of years to recognize patterns. This seemingly easy task turns into an extremely difficult feat when trying to express it through a computer algorithm. Neural Networks is one such methodology of expressing it. Like the way human learning process works, neural networks are designed to classify data based on recognized patterns in the data. For example, when given a large set of data, the neural network can identify a shape with loop on top of it as number 9. But number 9 can be written in different ways and not all could be identified by the network. This is where human training comes in. We provide the 'expert' advise and classify some shapes for the algorithm which helps it get more efficient and accurate.
The Deep Learning is concept of breaking down a larger set of problem into many smaller and easily solvable problem. For example, let's try solving the face detectection problem. Given picture how could you say that it represents a face? We can say that as long as it contains two eyes, a nose and lips it should be a face. So we are now breaking larger problem of face detection to smaller problem of identifying individual elements. This is where the Deep Learning comes in picture. It comprises of a "neural network" that has multiple layers of "neurons" all working towards solving granular elements of larger problem and being more abstracted from previous layers.
Usage And Applications
- Entertainment - If you are using Amazon Prime, you would recognize that it provides the ability to 'spot' actors in the given movie scene. This is the result of deep learning. Similarly, all the real-time movies, music, video suggestions clubbed with your consuming pattern are some of its implementations.
- Smart Homes - Have you though how Alexa, Siri, Cortana, Google Home recognizes your verbal command and executes an operation? That again the result of advancement in Speech Recognition techniques clubbed with deep learning.
- Autonomous Driving - Perhaps the best example, combining data from maps, weather, real-time sensor inputs (signals, other cars, animals, pedestrians, etc.) to make decision of slowing down or steering in a specific direction.
- Medical - Disease identification and diagnosis like melanoma. Clinical trial and research, etc.
After buying the new iPhone and being mesmerized by Apple Music, I felt an urgent urge to have a wir...
June 8, 2018 Google recently on their blog published a new feature to let users report spam on thei...
Numerous of our friends have been asking us recently whether should they be buying iPhone SE or iP...
Often times we hear about how GPU And Deep Learning are tied together and that later performs bett...