Machine Learning Fundamentals and definitions
Now after covering the mathematical fundamentals for machine learning Linear Algebra for machine learning — Part1,Linear Algebra for machine learning — Part 2, Calculus for machine learning,Probability for machine learning — Part 1,Probability for machine learning — Part 2
it’s obviously the time to start machine learning concepts and I am too very excited,but before jumping to learn machine learning concepts and algorithms,we need to understand few terms related to machine learning and data.I am not a guy much interested in these philophical definitions and terms,so It is short blog just written to make you feel comfortable in discussions ahead.
We will be focussing on classical machine learning terms in today’s context.
Machine Learning is defined as the use and development of computer systems that are able to learn and adapt without following explicit instructions, by using algorithms and statistical models to analyse and draw inferences from patterns in data.
Basically machine learning is the way of developing computer systems such that it can take decisions and adapt as per the requirement without prior instruction or explicit programming but by learning from previous data,observing it’s patterns and understanding it’s relations.
Machine Learning algorithms are mathematically designed algorihms that take the data,find and learn the mathematical pattern/relation in them such that they are able to give output/take decisions when operated on unseen data having similar relations between them.
The patterns or relations between data are generalized/derived many times and mathematical formulation having such relation is designed and later converted to learnable algorithm,which is later used in those kind of data(classible machine learning) and sometimes both pattern recognitions and formulation are dependent on machine if suffient data is available(self leanring/deep learning).
Based on the need of machine learning,machine learning algorithms can be classified into four typed:-
Supervised Learning algorithmns
When the data is labelled i.e data is in input-output form and machine learning algorithm works to take input data,observe patterns,do calculation,computation and then give output,it is called supervised learning algorithm.
example :- house price prediction data
The machine learning algorithm taking all other features of data like floor space,rooms,lot size etc.. and predicting/outputting house price will be a supervised learning algorithm.
Cancer prediction dataset
Machine Leanring algorithm taking a1,a2 ….. a9 as input and giving a10(B or M ) as output(B — no cancer,M-cancer) is also as supervised learning algorithm.
Types of supervised learning
- Classification — When the output lies in a discrete mathematical(finite) range and signifies the category/class of the data,it is called classification.for eg:- In cancer prediction,the output B/M represents class (cancer or no cancer),similarly in email spam filtering,the output gives class (spam/no spam),animal prediction from picture output also signifies a class of animal(dog/cat/bear etc..).
- Regression — When the output lies in infinite continuous range,and can be anything,for example :- price of house in house prediction,stock market profit prediction etc..,it is called regression
Unsupervised Machine Learning Algorithm
When data is unlabelled i.e input output form of data is not present,the machine learning algorithms analyses the features of data by observing there pattern and relation and tries to describe the structure of data by grouping/clustering the data points into sets etc..,these algorithms are called unsupervised machine learning algorithmns.
example:- lets consider the classification based cancer data example
Let’s remove d/a10 feature for a while,and suppose our dataset is unlabelled dataset with a1,a2….a9 i.e total 9 features.The unsupervised machine learning algorithm can learn pattern based on these features from data and then segment/group patients of similar pattern,similarity between there reports etc..
moreover,segmentation of customer by companies on basis of there reviews is also the example of unsupervised ML.
Semi-supervised Machine Learning Algorithm
The machine learning algorithms which observe the patterns and relations from labelled input-output based data but are used to describe the unlabelled data having similar patterns like similarity based grouping or dimension reduction etc.. are called semi-supervised machine learning algorithm.
Reinforcement Learning
Reinforcement learning focuses on regimented learning processes, where a machine learning algorithm is provided with a set of actions, parameters and end values. By defining the rules, the machine learning algorithm then tries to explore different options and possibilities, monitoring and evaluating each result to determine which one is optimal. Reinforcement learning teaches the machine trial and error. It learns from past experiences and begins to adapt its approach in response to the situation to achieve the best possible result.
We will learn them in detail when we learn ML algorithms from in depth.
Lets understand some mathematical persective of supervised machine learning algorithm as we will be covering it next.
Let’s consider the datasets provided above:-
House Prediction
Let Floor Space be x1,rooms -> x2,lot size -> x3 and Price be Y,so , an array X=[x1,x2,x3] denoting floor space,rooms,lot size
Suppose F is a mathematical function which takes X array as input and gives Y as output,F can be any mathematical function that suits to correctly predict the value of Y i.e => F([X]) = Y
F will have the parameters on which the relation between x1,x2,x3 and Y depends,for example let’s assume F is Linear Equation function,so,
F([X]) = F([x1,x2,x3]) => w1x1 + w2x2 + w3x3 = Y
Here the parameters are w1,w2,w3 . If data follows this linear pattern i.e if the price of house(Y) is linearly dependent on floor space(x1),rooms(x2) and lot size(x3),we have to only determine the correct values of parameters to find the correct functions.
If we computationally develop program to solve or learn the parameters of mathematical function provided data and type of function,it is called machine learning and that particular function is called machine learning algorithm.
For example in this case,Machine learning with some learning technique and given data,can learn the value of w1,w2,w3 and can predict the price of house given x1,x2,x3 in future.
Here also,there are different scenarios,we can predict the mathematical function observing data and then let machine learn the parameters,like in classical machine learning.
Or,we can let machine predict the mathematical relation or pattern in data and learn the parameters on that basis,for eg- self supervised learning,deep learning.
Well there are endless discussions and none will limit to perfect defintion of this topic.
For now and coming blogs on supervised learning,consider machine learning algorithms as the function F that takes features of data(X) as input,learns the parameters with several computations that will be discussed next,and gives the output F([X]).
Let’s move to our first machine learning algorithm,Linear Regression now with Linear Regression-Part 1 now………
Thanks for reading this,kindly read prerequisite mathematics for machine learning blogs before moving to machine learning.
Linear Algebra for machine learning — Part1,Linear Algebra for machine learning — Part 2, Calculus for machine learning,Probability for machine learning — Part 1,Probability for machine learning — Part 2
In case of any doubt ,or suggestion,ping me on +918292098293,please follow me on twitter https://twitter.com/AdityaR71244890?s=08 and linkedin https://www.linkedin.com/in/aditya-raj-553322197
