Blog‎ > ‎Machine Learning‎ > ‎

Machine Learning Explained in 2 Minutes!! [Part 3]

posted Jul 31, 2017, 11:39 AM by Rohit Bhaskar   [ updated Jul 10, 2018, 11:52 AM by Chirag Trasikar ]

The last 2 parts explained to you the different types of machine learning viz. Supervised and Unsupervised learning. This post will tell you about the types of machine learning, that are used in the industry, which differ on the basis of whether the algorithm is iterative or not.

In a non-iterative algorithm the trained model (i.e the thing that the machine learning code compares with while segregating the given input in Supervised learning) does not change until the user decides to create a new model. This means that every new input is always compared with the same model (set of guidelines) for the purpose of segregation.
The 2nd type is the iterative type of machine learning algorithm. Here the machine learning code changes our adapts with every new input that it is given to it.

The example of the 1st category is called as SVM (support vector machines). In this category, a trained model is made 1st using a database, and then this model is used for comparison and segregation of new inputs. The model once created never changes.
The example of the 2nd category is a method called neutral nets (neutral networks). In this method, a model is created in the same way as the SVM, but the difference is that that model changes whenever a new input is given for comparison.

Consider for example that we are segregating images of red cars and those containing other coloured cars. In either case (SVM’s or Neural Nets) we have to 1st create a trained model by providing our algorithm a set of images of red cars and a set of images of other coloured cars. Once the trained model is created, we can give it an input image of either a red car or another coloured car, and the algorithm will be able to segregate it.

Now, in SVM’s the algorithm will never change the model. But in Neural Nets, the model is changed on the basis of the input. For example, if I provide the algorithm with an image of a ‘light red’ coloured car, it will segregate it into the category of red car and then change the ‘mean’ slightly for the next comparison. i.e it is like it learns from every input that is is provided.

Take a look at the vid below to understand it better


I know that this post was slightly long, but i hope you understood it!! Comment below if you want any topic to be explained!