ML is a sub-part or can say an application of AI and it makes us understand the mathematics, logic of AI. It implements the technique of algorithms that allows computer programs to learn from the input data and automatically improve its results through the experiences without the need for any explicit programming. It mainly focuses on algorithms to learn from the data that is provided, collected insights and makes decisions that are based on previously unexecuted data by considering the information gathered.
We can perform ML using multiple approaches:
What are the algorithms that we use in building a model?
Algorithms are the list of instructions that we use to perform in building a model. These are implemented in such a manner so as to improve the quality of output. In algorithms, models are created with the ability to learn from the data which has been changed by the feedback system again and again. Some of the most used algorithms are Linear Regression, Logistic Regression, kNN, Decision Tree, Naive Bayes, etc.
Now, what is a model?
A machine learning model is trained on the training datasets which usually recognize the pattern in the data. So as to predict the outcome of the data that we have not inputted yet by the use of that pattern that is identified during the training of the data initially.
Machine learning is making computers more similar to human beings. Previously, the computers were coded to perform the actions that what to execute and what information are to retain or provide. This type of knowledge usually contains anything that can easily be trapped or captured. Nowadays, computers are encoded with the type of knowledge that is implied without being stated properly. It makes the computers capable to predict the outcomes itself after getting optimized to provide its best outcome. The knowledge that is given to computers is pronounced as tacit knowledge in which the knowledge is not stated properly but implements well as it comes out from the human and machine intelligence. Now, there is no need to worry about coding billion or trillion lines of code to perform calculations or any other task.
Let's take an example, working on the weather forecasting datasets we train a model as if either it will rain tomorrow or not by considering multiple factors. Here our model identifies a pattern to return either 1 or 0 means will rain or will not rain respectively. And algorithm here was found to be Logistic Regression.
Skills required: To reach our goal we need to clear our basics. So, to code for machine learning, we needed Python programming(can learn from this website), python libraries(NumPy, Pandas, scikit learn, matplotlib) in detail, class 12th basic mathematics, evaluation skills, and computer fundamentals.
Deep learning is a subpart of Machine Learning and it is one of the ways to achieve ML.
ML fails to achieve better accuracy at some point even though you give more data.
Deep Learning can provide better accuracy by making use of neural networks because it is hungry for data it performs well (you need to take care of overfitting).
A neural network is a set of a task-specific algorithm that makes use of a deep neural network that is specifically inspired by the structure and function of the human brain.
feature extraction is automatic in deep learning.
-- 'TRAIN ME'
-- 'I AM SELF SUFFICIENT IN LEARNING'
-- 'MY LIFE, MY RULES'
Supervised Learning is the one, where a supervisor is needed to perform a certain task so that machines can learn easily. We have a data set that acts as a teacher and its role is to train the model or the machine. Once the model gets trained it can start making a prediction or decision when new data is given to it.e.g-Linear regression, Logistic regression, kNN.
What are the uses of Supervised Learning?
This type of learning is mainly used in spam filtering, language detection, computer vision, search, and classification based problems.
Here no such teacher is provided, only certain data is provided and in its contrast, there is no output. Hence machine learns through observations(algorithms) and implements it as experiments, on its way to find the best result. It clusters its data but can't be able to add labels to the cluster like it cannot say this a group of apples or mangoes, but it will differentiate all the apples from mangoes. K-means clustering is an example of unsupervised learning.
What are the uses of Unsupervised learning?
This type of learning is basically used in the segmentation of data, anomaly detection, recommendation systems, risk management, fake image analysis.
Here also the aim of our algorithm remains the same to have the best output during interaction with the environment. It follows the concept of the hit and trial method. Our model is rewarded or penalized according to a correct or a wrong answer, and on the basis of the positive reward points gained the model trains itself. And again once trained it gets ready to predict the new data presented to it.
What are the uses of Unsupervised learning?
This type of learning is used by self-driving cars (or autonomous vehicles), games, robots, resource management.
Nowadays, Machine Learning is being adopted by many industries across the world to deal with huge amounts of data. Being aware of the leveraging insights that is obtained from the data, numerous companies are able to work in a more efficient manner to compete with the other companies in the same field. ML helps in the progress of many fields. Some of the applications of the Machine Learning are:
Web search: It includes selecting and ranking pages based on what you are most likely to have.
Computational biology: It is rational designing of drugs on the computer which is completely based on past experiments and experiences.
Finance: Its aim is to decide whom to send and what credit card offers to. It also includes the valuation of risks on credit offers by verifying the security measures. It helps in predicting the best possible regions and ideas of investing money.
E-commerce: It includes the prediction of customer churn whether the transaction is fraudulent or not.
Space exploration: Radio astronomy and space probes.
Robotics: How to handle an unfixed new environment and how to perform the tasks with respect to it. A self-driving car (having an ability to work from the learned experiences).
Information extraction: Globally, it asks questions over databases.
Social networks: Machine learning to extract the outcomes from data.
Debugging: It is frequently used in CS problems such as debugging. It is a part of the labor-intensive process and suggests where the bug could be.
Healthcare: ML helps in knowing the overall health condition of real-time patients such as blood pressure, heartbeat, or any other vital parameter by analyzing his conditions faster and also let the patient aware of the occurrence of any disease in the future just by studying the patient's history records.
Hey!!, Have you learned all concepts?
If yes, then you know you are now able to code for Machine Learning.