A JUNGLE BOOK ON MACHINE LEARNING
Machine learning stories are no more a subject of science fiction stories. Wide range of applications starting from genetics to business process optimization, generated the required momentum wherein people coming from different backgrounds have started taking interest in machine learning and it has become an important constituent of Data Science.
WHAT IS MACHINE LEARNING?
Machine learning is the technique to make your computer smart so that it can act accordingly and for that you don’t need to program it explicitly. It has three prerequisites – data availability, statistical method or algorithm and use of computational power. The learning process consists of three steps:
- Data : Recording of observations
- Abstraction : Applying logic into data
- Generalization : Use of abstraction to get insights.
Application in statistics and computational complexity which come automatically with it, stopped many to go beyond headlines. In this situation, what if you have a story book, telling you stories that make you understand complex machine learning algorithms and application in data science! Let us see how the story goes!
STORY OF AN ASHRAMA IN A FOREST (not random!!)
The story starts with an ashram in the jungle, famous as Acharyamuni’s ashram. He was famous for his unique teaching technique. He had many students from different parts of the country. After many years of teaching, he decided to train one of his best students his unique method of teaching so as to transfer his learning. One day, he called his student Arya and told him to sit beside him while he was going to teach some new students.
Next day Acharya started teaching a student, Bargi, who has joined the ashram recently. Acharya said “I will give you some task today, and I will send someone to guide you. Tomorrow you need to do the same task on your own.” Acharya then called a person named Rajan. He asked Rajan to guide Bargi to the jungle to identify friendly animals.
Next day, Rajan guides Bargi to identify friendly animals one by one. At the very beginning, Bargi notices some monkeys. Rajan placed some fruits in front of the monkeys. They immediately took it. Rajan said “monkeys are quite friendly to us for years”. They moved on, and after some time they encountered a deer. Rajan tried to give it some fruits, but the deer did not come close, and hid in the bushes. Rajan said with a smile “they don’t come close to human but do not harm either.” They continued their journey and after a while they noticed a fox, Rajan told Bargi very firmly “stay away from this animal as much as you can.” Bargi nodded his head with some amount of confusion in his mind. After that they returned to the ashram. Next day Bargi had to take the journey on his own.
Early in the morning Bargi started his journey. At first, he encountered a dog, and tried to give him some fruits. The dog did not take it. After that, he noticed a bird, and again he tried to offer some fruits, but the bird flew away. A rabbit, present nearby, ate the fruit. Bargi was in doubt as to what he would say to his Acharya. He came back to his ashram. In the evening, Acharya called him and asked what he learnt that day. Bargi shared the experience and classified dog and bird as unfriendly and rabbit as friendly. Acharya smiled and said “not bad as a beginner, but someone needs to explain it further to you “.
SUPERVISED MACHINE LEARNING
By now, you have understood the basic concept of ‘supervised machine learning’ where you need to train your machine to construct some logic from the past events and make predictions on new data. If you recall, when Rajan takes Bargi to the jungle to identify friendly animals, he tried to give fruits to different animals and Bargi noticed and judged every single incident, based on how they were reacting when fruits were given to them. In technical term, this is called ‘learning from training data’ i.e. machine learns how data is performing in different circumstances. This is a very important aspect of supervised machine learning algorithms in data science.
However, Bargi could not classify dog and bird as friendly animals (who are actually friendly in nature) when he made the visit alone. In technical term, we call it performance on the test data, where logic which has emerged from the training data is applied on test set to check the accuracy of the model built. Since this was the only criterion of selection, Bargi could not perform well. In technical terms, we call it ‘accuracy’, where there is a mismatch between actual and predicted outcome. Performance of Bargi could have improved if Rajan had used more criteria of selection, the technical term for it is ‘improving model performance’. There are several ways to improve the model, increasing the number of criteria is one of them.
If you would like to know more about the courses offered on data science and machine learning, visit the website of Data Brio Academy.