Why Machine Learning ?
I always love to start my blogs with explaining "why" we are even talking about that topic. The reason behind this is that i strongly believe that the "why" should always be clear to us whenever we are heading towards something new. When we have the answer of "what is it?" we basically just "know" the topic but when we have the answer of "why this?" we start understanding the topic. You are maybe thinking that why am i discussing all of these philosophical stuff but these words are for those who will become learners from just readers of this blog.
Getting back to the point, now we will discuss that why Machine Learning had been introduced to human. In typical words Machine Learning is the ability of computational machines to imitate human-like behavior/intelligence. What type of human intelligence we are talking referring to? We are talking about human's ability of recognizing hidden patterns and predicting things. We can teach a machine how to unleash hidden patterns and make predictions based on the patterns. This is basic concept of Machine Learning. Our traditional computation system does not allow the machines to learn from given data by their own but ML allows us to do so. Application of ML in various fields like business, Education, Research etc. has unleashed many hidden possibilities.
Machine Learning Categories
There are three ways of a machine to learn from data, discover the hidden patterns and unleash the story behind the dataset. When we train a machine learning model using labelled data which means we tell the machine what the input dataset is about and what to predict is called Supervised Learning. The second way of training an ML model is using un-labelled dataset which means we won't tell the model what is the input dataset about and what to predict. Is is sounding absurd? okay, let me explain this a bit more. Suppose we are giving the medical history of the cardiac patients a doctor has seen in a year as the input dataset to train an ML model. We are not telling the model what we actually want the model to predict. Suppose maximum of the patients who were detected to have severe cardiac issue had diabetes, so if we give the medical history of a new patient who visited to the cardiologist the ML model will be able to predict that whether that person has a severe cardiac issue or not before the physical check-up.
The third way an ML model can be trained is through reward or punishment based learning which is called Reinforcement Learning. In this method the model will get rewarded(in the form of positive feedback) for producing the desired behaviors and will get punished(in the form of negative feedback) for undesired behaviors.
Top comments (0)