Question 1: What is Machine Learning ?
Answer: Machine Learning is a subset of artificial intelligence. In simple terms we study computer algorithms in machine learning and performance of these algorithms depends on type and amount of data (used to train a machine learning model). And these algorithms learn with past experiences which help in improvement of the performance of a computer program like human being.
Example: In robotics, to make future prediction based on available data, self driving car etc...
Question 2: Types of Machine Learning ?
Answer: When we are talking about types of machine learning, we are basically talking about different ways in which a machine can learn.
Basically there are three ways in which machine learn:
- Supervised machine learning
- Unsupervised machine learning
- Reinforcement machine learning
Question 3: How these three type of machine learning algorithms differ each other ?
|Supervised Learning ||Unsupervised Learning ||Reinforcement Learning |
|Machine learn with labelled data under external supervision. For example; teaching a child what is a apple and what is banana.||Machine learn with unlabelled data under minimum supervision . for example; Let's take a case where no one to tell a child what is the difference between an apple and a banana ,In this case, he will find out by looking at the patterns of things himself and will teach himself by recognising pattern.||This the most advance learning in machine learning . In this case an agent interact with unknown environment and learn through trials and errors using previous feedbacks. For example self driving car, in NLP.|
A simple example; In ancient time evolution of human from an animal where he learn from his trial and punishment method.
|Problems related to Regression and Classification ||Problems Related to association and clustering ||Negative and Positive Reinforce |
Note: It is true that reinforcement learning is best out of three but it is difficult to deal with reinforcement learning in practical. Or learn it when you mastered the basic learnings.
Further Classification of supervised and unsupervised machine learning:
Question 4: Python Libraries need to be know to start with Machine Learning ?
Pandas : Pandas full tutorial
Python Pandas library is a fast , flexible and easy to use library to manipulate ,merging, sorting data for data analysis.
Numpy: Numpy Tutorial
Numpy library in python deals with all mathematical and scientific computing include all matrix/ array calculations.
Matplotlib: Matplotlib tutorial
Matplotlib library is a plotting library in python. We can make static plots using maplotlib but not interactive.
Advance plotting libraries: Plotly , Seaborn , Folium Library , Geopandas (geospatial data)
Scipy: Scipy Library tutorial
Scipy in a computation library that uses NumPy. Basically advance form of Numpy.
scikit or sklearn :
Scikit-learn (known as sklearn) is a free software machine learning library for the Python programming language. Basically we can directly use some features of machine learning (for example: regression , classification etc..) and also can download some famous datasets using sklearn directly.