Starting with the basics of machine learning in R

Deep learning is a subcategory of machine learning inspired by the structure and functioning of a human brain. In recent times, deep learning has gained a lot of traction primarily because of higher computational power, bigger datasets, and better algorithms with (artificial) intelligent learning abilities and more inquisitiveness for data-driven insights. Before we get into the details of deep learning, let's understand some basic concepts of machine learning that form the basis for most analytical solutions.

Machine learning is an arena of developing algorithms with the ability to mine natural patterns from the data such that better decisions are made using predictive insights. These insights are pertinent across a horizon of real-world applications, from medical diagnosis (using computational biology) to real-time stock trading (using computational finance), from weather forecasting to natural language processing, from predictive maintenance (in automation and manufacturing) to prescriptive recommendations (in e-commerce and e-retail), and so on.

The following figure elucidates two primary techniques of machine learning; namely, supervised learning and unsupervised learning:

Supervised learning: Supervised learning is a form of evidence-based learning. The evidence is the known outcome for a given input and is in turn used to train the predictive model. Models are further classified into regression and classification, based on the outcome data type. In the former, the outcome is continuous, and in the latter the outcome is discrete. Stock trading and weather forecasting are some widely used applications of regression models, and span detection, speech recognition, and image classification are some widely used applications of classification models.

Some algorithms for regression are linear regression, Generalized Linear Models (GLM), Support Vector Regression (SVR), neural networks, decision trees, and so on; in classification, we have logistic regression, Support Vector Machines (SVM), Linear discriminant analysis (LDA), Naive Bayes, nearest neighbors, and so on.

Semi-supervised learning: Semi-supervised learning is a class of supervised learning using unsupervised techniques. This technique is very useful in scenarios where the cost of labeling an entire dataset is highly impractical against the cost of acquiring and analyzing unlabeled data.

Unsupervised learning: As the name suggests, learning from data with no outcome (or supervision) is called unsupervised learning. It is a form of inferential learning based on hidden patterns and intrinsic groups in the given data. Its applications include market pattern recognition, genetic clustering, and so on.

Some widely used clustering algorithms are k-means, hierarchical, k-medoids, Fuzzy C-means, hidden markov, neural networks, and many more.