Module 1#

Machine Learning#

A scientific discipline that explores the construction and study of algos that can learn from data
Such algos operate by building a model based on inputs and using that to make predictions or decisions rather than following only explicitly programmed instructions without human intervention

Traditional Programming VS Machine Learning#

We give the computer some data and a program and we expect an output from the same in the case of traditional programming
Pasted image 20220116112918.png

In case of ML we give the computer some training data and some expected output and we end up getting a procedure that solves the problem for new input data
Pasted image 20220116112930.png

Definition by Tom Mitchell (1998)#

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 task in \(T\), as measured by \(P\), improves with experience \(E\)

Where does ML fit in?#

Pasted image 20220116113818.png

Steps of ML#

Define Business Problem
Gathering Data
Prepare said data (Takes most of the time, around 80%)
Choosing a model
Training
Evaluation
Hyperparameter Tuning
Prediction

Types of Learning#

Supervised/Inductive Learning
Unsupervised Learning
Semi-supervised Learning
Reinforcement Learning

Supervised Learning#

Pasted image 20220116120137.png
This learning type uses data that is labeled. Data that have categorized data into columns are considered labeled data.

In the above image we are trying to predict a genralized function \(f(x)\) that will predict the potential value of y given a new value of x.

Regression#

\[ y = f(x_1, x_2, x_3,..., x_n) \]

\(y = Output\)

\(f() = Prediction Function\)

\(x_1, x_2,.., x_n = Features Used\)

Training: Given a training set of labeled examples, estimate the prediction function f by minimizing the prediction error on the training set
Testing: Apply f to a never before seen test example x and output the predicted value y f(x)

Classification#

Pasted image 20220116123809.png

Unsupervised Learning#

Pasted image 20220116124304.png
Unsupervised learning helps in getting to know not so obvious attributes that categorize data sets into smaller groups in case of clustering

Reinforcement Learning#

No pre-defined data
Semi supervised learning model in ML
Allow an agent to take actions and interact with an environment so as to maximize the total rewards
Examples
- Autonomous Cars
- Game playing
- Robot in a maze

Difference between the three types of learning#

Pasted image 20220116125417.png

Other Categories of Learning#

Pasted image 20220116125550.png
- In batch learning there is a potential for relearning when an updation is needed. It is used ideally done once before deployment
- In case of changing data sets we use online/incremental learning

Pasted image 20220116125858.png
- Instance based learning: also called as lazy learning. Learning techniques do not build a model but stores all training instance in memory and when they undergo classification they use proximity measures k closely related members to categorize them.