From Here and there, everybody is shouting about Large Language Models, Generative AI, and ChatGPT. The root of these shouts come from so called Artificial Intelligence (AI). In few words, AI includes machine learning, machine reasoning, and robotics. Machine Learning is comprised of models such as deep learning and reinforcement learning. Machine Reasoning brings planning, knowledge representation, search and optimization. Whereas, Robotics covers control, perception, sensors and actuators, briefly cyber-physical systems. Out of all these information, you might be thinking that where should I start to learn AI. To let you understand the broader field of AI, I am going to start with a very crucial component and it is Machine Learning Models.
Let’s dig deep while staying at a surface of this big world of AI!
What are Machine Learning Models?
Machine Learning Models are basically computational algorithms. These algorithms are made in such a way that they can learn specific patterns from data. As a result, they can give us predictions without being explicitly programmed for a specific task. But wait! We forgot to define Machine Learning. Let’s pull over our car to this point.
Machine learning, the sub-field of artificial intelligence, is a technique for making computers learn patterns and meaningful relationships through experience from observations and examples. Machine learning is evolving at a very fast pace with ongoing research and development. It is capable of bringing automation of complex tasks, and making your world full of intelligent systems. The algorithms of machine learning (as known as machine learning models) are classified as supervised learning algorithms, unsupervised learning algorithms, semi-supervised learning algorithms, and reinforcement learning.
We did all Bla, Bla, Bla. and now you might be wondering that where am I going to use machine learning models. To give a little glimpse, let’s talk about area of application. Machine Learning Models are applied in areas such as recommendation systems, natural language processing, image recognition, autonomous vehicles, finance and many more. However, the result of a system depends on the specific task and the nature of the data available. Now, it is time to get a deep dive in machine learning algorithms.
What are types of Machine Learning Models?
1) Supervised Learning:
Supervised Learning is a model training in a supervised manner. Meaning, in order to train a model in a supervised manner, you need a labeled dataset. This means that input training data has to be made of set of input and associated actual labels. In this type of learning, the ultimate goal of training the model is to learn the relationship between inputs and outputs. As a result, a trained model can make as accurate predictions as possible on new and unseen data.
The common supervised tasks are classification and regression. Classification classifies the data to particular class, whereas regression build a function in which data can be fit as accurate as possible. Most popular algorithms of supervised learning include linear regression, decision trees, support vector machines, and neural networks.
2) Unsupervised Learning:
We just looked into supervised type of learning, where we are in need of labeled dataset. What if we do not have access of labeled data, or it is too hard to get it and prepare it. There comes unsupervised learning into a picture. In structured words, Model training on an unlabeled dataset is Unsupervised Learning. Here, the model looks for patterns, relationships and structures in the data without any guidance on the output explicitly.
The common tasks of unsupervised learning are clustering and dimensionality reduction. Most popular algorithms of unsupervised learning include K-Means clustering, hierarchical clustering, and principal component analysis (PCA).
3) Reinforcement Learning:
This type of learning generates an optimal solution in a specific environment with help of the agents and reward function assigned to them. Most popular algorithms of reinforcement learning include Policy gradient methods, Q-Learning, deep Q Networks (DQN).
4) Semi-Supervised and Self-Supervised Learning:
Semi-Supervised learning is a hybridization of supervised and unsupervised learning, where training is done on labeled data and unlabeled data as well. On the other hand, Self-Supervised learning is one type of unsupervised learning. But, here, instead of not having labels completely, the model generates its own labels from the input data.
Conclusion
Welcome to the conclusion. After visiting machine learning models and its types, a question that is meant to be aroused in your mind is which model and when should it be used. This totally depends on the specific task and the nature of the data available. Moreover, factors such as interpretability, scalability, and computational efficiency have to be considered while choosing a machine learning algorithm or model for a specific application.
