Basics of Machine learning

Published in

AlmaBetter

6 min readJul 20, 2021

Breifing about AI,ML,DL and DS .

Nowadays we hear these words more often such as Artificial Intelligence, Machine Learning, Deep Learning, and Data Science. Out of 10 people, I hear these words from 7–8 of them, and why not Data Science field is booming, and it has been considered as the sexiest job of 21st century. So, it’s obvious to hear those terms. But I see out of that 7–8 people, only 2–3 of them know the difference between Artificial Intelligence, Machine Learning, Deep Learning, and Data Science. So let's just go through each term and understand the difference between them and also later we will see Machine learning in-depth.

Artificial Intelligence

Artificial Intelligence is computer programs that learn to perform tasks that require types of Intelligence that are usually performed by Humans. Basically, it tries to mimic Human Behaviour. According to the situation, the machine can take its own decision without human intervention. While creating AI app sometimes use Machine learning and deep learning within them.

Application of Artificial Intelligence

a)SELF DRIVEN CARS

b)Roomba -Roomba is a compact, computerized vacuum cleaner that automatically guides itself around your home. The latest Roomba, the 980 model uses true AI.

c)CHATBOT, ALEXA, SIRI

Machine Learning

It is a subfield of AI. It is the study that gives machines the ability to learn without being explicitly programmed. Basically, there is no exact definition of machine learning.ML gives us the statistical tools to deep-dive into the data.

Tom Mitchell defined ML as :

“A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.”

let's understand the above definition using an Example :

Suppose our Email program watches us which email we are marking or not marking as spam and based on that learning we will try to define E, P, and T. So, here the T(Task) Would be Classifying Emails as spam or not spam, E(Experience) would be watching us label emails as spam or not spam. and finally, the P(Performance) would be the number of emails correctly classified as spam /not spam.

Application of ML

a)Weather prediction

b)Handwriting recognition

c)Autonomous helicopter

Deep Learning

It is a subset of AI and ML it tries to mimic the human brain via a neural network. It requires a huge amount of DATA and Computational Power.

it has many techniques such as :

a)ANN-Artificial Neural Network

b)CNN-Convolutional neural network

c)RNN-Recurrent Neural Network

Applications:

Voice Recognition
Video Recognition
Text Analytics

Data Science

Data science not only comprises of AI, ML, DL but there is also one thing which is very essential if anyone wants to survive in the DS field is having ‘Business Knowledge’.Data science is a field that combines statistics, data analysis, scientific methods.

Data science’s primarily objective is to fetch meaningful information and insights from Data. That is only possible with the help of various mathematical and Machine Learning-based algorithms. Hence, Machine Learning is a key element of Data Science. Some of the types of data used in DS are image data, text data, video data, audio data, etc.

Let us see more about Machine Learning

Machine Learning has different techniques such as Supervised Learning, Unsupervised Learning, and Reinforcement Learning.

Supervised Learning

As the name suggests Supervised means nothing but being under the supervision of someone (Given the feature and target variables with output).In this, we have to teach the machine first how to perform the task, and then after several experiments machine performs the tasks by itself by observing the experience.

we have two types of Supervised Learning

Regression
Classification

Regression

It Predicts the continuous-valued output. That is, the output of the function is not categories, but is a continuous value.

Classification

It Predicts the discrete-valued Output (0 or 1 ,yes or no,etc).This is a process where the dataset is categorized into discrete values or categories.

Example

a) Regression - Given data set consists of weight and height try to predict the BMI for each weight and height data.BMI as a function of weight and height is a continuous output, so this is a regression problem, Future price of the stock market, predicting weather forecasting, etc

(b) Classification — Given a patient with a tumor, we have to predict whether the tumor is malignant or benign, Predicting monsoon will be normal next year, etc

Unsupervised Learning

These are called unsupervised learning because there is no one to supervise. In unsupervised learning we don't have any labels, it approaches problems with little or no idea about how our output might look like.

Unsupervised learning problems can be further grouped into clustering and association problems.

Clustering: A clustering problem is where you want to discover the inherent groupings in the data. In this algorithm, we form different groups of the same feature(Same characteristics) where each group will be called a cluster.

Application of Clustering :

Google news — if a story or news is related or of the same context, it groups and shows it in one cluster. As we can see in the picture there are different stories but because of the same category, it's mentioned in one cluster. (There are 2 clusters shown in the picture)

2.Social network analysis

3.Market Segmentation,etc

Non Clustering: In this Algorithm, it transforms or Alters the input/data. Let’s understand this by taking the famous example of the cocktail party algorithm

Application

A common non-clustering problem would be the cocktail party algorithm. Suppose that two people and two microphones are recording at two different locations at a gathering. Mic 1 is closer to Person 1, and Mic 2 is closer to Person 2. Both mics can devour the conversation, however, the voice of the closest person is louder. After putting the recording through the party Algorithm, the outputs would have separated the 2 voices. It allows you to seek out structure during a chaotic environment, and it’s transforming the info. Therefore this is able to be a non-clustering algorithm.

Reinforcement Learning

This learning algorithm is based on Reward and penalty policy. This is mainly inspired by behaviorist psychology. Suppose we have an agent in some environment. An agent performs any task in that environment and based on his performance we will either give reward or penalize. And based on this learning its accuracy increases.