By- Vineet Maheshwari
To infinity and beyond.

0 likes followers Views

Vineet Maheshwari


Supervised Learning and Unsupervised Learning

Hey !

In previous article we learnt about different Machine learning techniques and now is the time to learn about 2 of those techniques in depth. So let's dive in.

Supervised Learning - Now before learning about what is supervised learning, first we need to build our understanding about labeled data. Labeled data means, answer is tagged along with the target data for which we will execute algorithms. For example, labeled data of dogs would tell algorithm to tell us about different breeds, whether they are of dogs or of some other animals. And whenever a new image will come up, it will find out the result based on the training dataset.

Now understand about supervised learning, it means learning under supervision of someone who can judge us whether we are doing this thing in a right way or not.

This type of learning helps in 2 types of problem, classification and regression.

Classification problems ask the algorithm to predict discrete values, identify input data as a member of particular class or group. For example, if there is a training data set of dogs, that means each image is pre-labelled with some sort of category. Now the algorithm is being tested to correctly classify the new images of dogs.

And on the other hand, the regression problems look for continuous data.

Therefore supervised learning is best suited to problems where there is a set of available reference points.

Unsupervised Learning: This technique is used where the model is being fed with the data set with no explicit instructions, that what to do with it. The model then tries to automatically find the structure in data by extracting the features and analyzing the structure.

It can organize the data in various ways like Clustering, Anomaly Detection, Association, Auto-Encoders.

Predicting accuracy in this case is quite difficult.

Now a little something about semi-supervised learning.

It contains both labeled and unlabeled data. This method has an advantage and can be used whenever it is difficult to extract features or it is taking too long to label the examples.

This kind of situation is mostly faced in medical field. But the deep learning network can still work in more accurate way by working on a small proportion of labeled data and improving the accuracy in comparison to unsupervised learning.

We will learn about the flow of machine learning in next article.

Till then, keep learning !

HelpFeaturesMade with in INDPrivacyAbout
© 2020