There are mostly four approaches are used in modeling for deep learning models. All of them are mainly separated because of different data present. These approaches are





We will try to explain each of them in some details with some appropriate example if required.



This technique is mostly used in data science (deep learning) to generate model now a days. In this technique we use labeled data to generate models. Now the natural question arise. What is labeled data? and how it is used?

Labeled data is a data which is labeled by human. Such as we have dog image which is seen by human and labeled it as dog. Labeled means change name of image as dog ). Likewise all these data which human first see it and then recognized and label it like Audio, Video, Image, Image pixels everything can be labeled.This labeled data can be used in training model to understand and learn that all of this type of data is in this category such as, let say we have 1000 images 500-500 for dogs and cats. Now we feed this data to train deep learning models. Training means showing that data to deep learning models. Due to which model will learn that all this type of images are dogs and cats. Then in future whenever you show any dog images to model it will understand that this is dog image.

For clarification model generally considered as a piece of software which has capabilities to take input.



You can consider model as a child to whom you are show different data, such as image and telling this is dog or cat.which in exchange will learns that this type of shape is of dog or cats.

This approaches is mainly used for classification of certain objects from data and for regression purpose. To predict some value based on deep learning.



This technique is mostly used to group data of similar kind. In this of approaches we passes unlabeled data to the deep learning model which groups data of similar fashion to one group which after the grouping it is labeled by human that is group of data . This idea is very fascinating but it is very difficult to achieve . KMEAN library is used for this kind of approaches



This technique is a combination of both techniques such as supervised and unsupervised leaning. Supervised learning is used to learn some representation and Semi-Supervised learning is used to regroup unseen data which is available in the dataset . Let say we have some data partly labeled data and partly unlabeled. Partly labeled data is used to train model a hit to classifies to some extend and then unlabeled data is used to group for that classification.

This also hard to develop and model.



In this type of technique model (Agent) learn through experience as human learn. In this type of approaches model is introduce to the environment in which agent (Model) performs certain steps. Those steps (Action) is already defined. If agent takes right step than it is reward with points else it is penalized due to which after every wrong action agent learn that all at this time step and in this situation I have to take this steps. Due to this approaches deep mind of google. Google have made alphago model which has beat human champion in Go match. They have also made models which have beaten chess master also.



Environment, Agents all are software which is used to train agents for future task.

It is mainly developed with the mythology as try and error learning.



In this post we have learning about different techniques while developing deep learning models. Through our observation we understand that labeled data is highly important. If you have got labeled data then generating deep learning model as really easy. But generating labeled is expansive. If you got free labeled dataset than you are very lucky.If you have any query or feedback or need any help regarding boiler code of any topic , you can freely ask in comments. I would really love to help you.



Taher Ali Badnawarwala

Taher Ali, drives to create something special, He loves swimming ,family and AI from depth of his heart . He loves to write and make videos about AI and its usage

Leave a Comment

No Comments Yet

Leave a Reply

Your email address will not be published. Required fields are marked *