I remember sitting in a machine learning course a few years ago. It was a sunny day and everyone was talking
about this course and how the theory that we learn in this course would be changing the future.
I am glad I took that course and now I am successfully working on interesting machine learning projects.
In this post, I would like to give you a brief introduction to machine learning and my anaylsis on the future of this field
Let’s start with the end goal of any machine learning project. Any machine learning project, from a business perspective, has three goals
- Automate and
Imagine all ther services that you are using on a day to day basis.
- Movie Recommendations
- Book Recommendations
- Visual Inspection
- Route optimization in the supply chain industry
- Prediction churn rate of your subscribers etc.
All of these trends show that the world is moving to a more automated and data-driven society.
So how does machine learning work behind the scene and what are the core algorithms used to make
these systems intelligent.
Machine Learning can be categorized under three main categories
- Supervised Learning
This is the most common use case of machine learning and the majority of industries still rely on Supervised
machine learning algorithms to drive their business forward. In very simple words, without getting too mathematical, a supervised learning algorithm consists of some input X with some labeled output Y and the task is to learn the function X->Y to predict an unseen input.
- Unsupervised Learning: In an unsupervised learning setup, there is no information of the output and the system needs to find patterns inside a dataset. In essence, this is not a trivial task and requires a lot of engineering. A very popular unsupervised algorithm is K-means clustering algorithm.
- Reinforcement Learning: The king of all algorithms, Reinforcement learning algorithms will see more acceptance in the near future. In essence, reinforcement learning algorithms work on the basis of a reward system. For every action taken by an entity, a reward is awarded. The task is then to optimize the actions so that the system gets a good reward for future actions. A popular use case of reinforcement learning is in demand forecasting in the retail industry.
Future of Machine Learning and AI
Although lot of companies have adopted AI and machine learning in their core business activities, most companies suffer from two
- Different Silos
- Lack of information architecture.
Companies are struggling to automate the process of collecting data, building machine learning models, analyzing data, and
maintaining machine learning models. Each of these tasks has different teams with different tools and libraries to work with.
In the future, we will see companies invest more time and money to work on a unified and simplified architecture to make sure companies are spending their time on the core business activity and not so much on building infrastructures for each and every task thereby increasing productivity and efficiency of data-scientists to help analyze and find meaningful insights from the data.