Supervised learning is the first concept of machine learning algorithms in the field of Artificial intelligence. In Supervised learning, input data(unlabelled data) is evaluated by an algorithm with the help of prior data that has been categorized as labeled by humans manually. After the development of the idea, there have been explorations on countless algorithms - to enhance the quality of output of machine learning and to minimize the manual input of unlabelled data to save both time and effort.
Semi Supervised Machine Learning is the concept of supervised learning where the algorithm uses a small amount of labeled data along with a larger amount of unlabelled data to predict the labels of newly Included unlabeled data. Conceptually it comes under both supervised learning and unsupervised learning. In a broad sense, it can be further explained in an elaborative way. let's take a model of algorithm which uses its previous knowledge to examine various data and then categorizing it into different classes it has learned in its previous experience. While doing so, it uses labeled and unlabeled data to perform certain learning tasks. These algorithms learn as it continuously improvising themselves to minimize the error. During the evaluation, the algorithm would choose between the supervised route and unsupervised routes. Supervised routes operate only on the data with labels, reducing the size of the dataset; otherwise, the latter eliminates the labels while keeping the rest of the dataset for cluster analysis.
A Semi Supervised Machine Learning algorithm uses labeled data to train itself, turning itself into a "partially trained model". later this partially trained model uses its knowledge to determine the class of unlabeled data and put them into specific categories, known as "pseudo labeled data". In the final stage "pseudo labeled" data and "partially labeled" datasets are combined, creating a unique algorithm that merges both the descriptive and predictive details of supervised and unsupervised learning.
One of the real-world examples of Semi-Supervised learning is a text document classifier. It is an ideal Situation where semi Supervised learning comes in handy because it is nearly impossible to find a large amount of labeled data to analyze the text document. That's why it's ideal to use a semi Supervised Machine Learning Algorithm.
Another good example of Semi Supervised Machine Learning in a photo archive, where only a few photos are labeled such as apple, banana, etc. and a majority are unlabeled.