Lec 04-Introduction to AI Algorithms

Summary

This video script offers an insightful overview of AI algorithms in marketing, focusing on their definition, complexity, and function. It delves into the three primary learning patterns: supervised, unsupervised, and reinforcement learning, each with its distinct set of algorithms like decision trees, random forests, and neural networks. The script also touches on the Algorithmic Bill of Rights, emphasizing the importance of awareness, accountability, and validation in AI algorithm usage.

Takeaways

• 🧠 AI algorithms are sets of instructions for computers to learn and operate independently, significantly more complex than general algorithms.
• 📚 The learning patterns for AI include supervised learning, unsupervised learning, and reinforcement learning, each with distinct training and functioning methods.
• 🌳 In supervised learning, algorithms are trained with labeled data to predict outcomes, akin to a student learning with a teacher's guidance.
• 🔍 Unsupervised learning uses unlabeled data to find patterns and relationships within the data, without any prior guidance.
• 🤖 Reinforcement learning algorithms learn from feedback in the form of rewards, improving actions based on the environment's responses.
• 📊 Common supervised learning algorithms include Decision Trees, Random Forests, Support Vector Machines (SVM), Naive Bayes, and Logistic Regression.
• 🔢 Linear Regression in supervised learning is used for continuous predictions, like sales forecasting, based on the relationship between variables.
• 👥 Unsupervised learning examples include K-means clustering for grouping data points and Gaussian Mixture Models for more complex cluster shapes.
• 🕵️‍♂️ K-Nearest Neighbors (KNN) is an algorithm used for both classification and anomaly detection, based on the proximity of data points.
• 🧬 Neural networks mimic the human brain's functions, with interconnected nodes organized in layers, capable of pattern recognition and complex tasks.
• 👮‍♂️ The Algorithmic Bill of Rights outlines principles for ethical AI use, emphasizing awareness, accountability, explanation, and validation to prevent biases and harm.

Q & A

What is an AI algorithm?

An AI algorithm is a set of instructions for a computer to learn and operate on its own. It is a complex programming that determines the steps and learning capabilities of an AI program.

How do AI algorithms work?

AI algorithms work by taking in training data, which can be labeled or unlabeled, and using that information to learn and grow. They complete tasks using the training data as a basis.

What are the different types of learning patterns in AI algorithms?

The different types of learning patterns in AI algorithms are supervised learning, unsupervised learning, and reinforcement learning.

What is supervised learning in AI?

Supervised learning is a category of AI algorithms that work by taking in clearly labeled data during training to learn and predict outcomes for other data.

Can you explain the decision tree algorithm in supervised learning?

A decision tree algorithm is a supervised learning method that classifies data into nodes, with a root node and leaf nodes, using attribute selection measures like entropy and information gain.

What is the random forest algorithm and how does it differ from a single decision tree?

The random forest algorithm is a collection of multiple decision trees that are used to gain more accurate results. It differs from a single decision tree by adding randomness to the model and considering the majority vote or average of multiple trees for the final output.

How does the support vector machine (SVM) algorithm work?

The support vector machine algorithm works by plotting data in an N-dimensional space and finding the hyperplane that best separates the classes. It aims to maximize the margin between the nearest points of different classes.

What is the role of the Naive Bayes algorithm in AI?

The Naive Bayes algorithm is a classification algorithm that assumes the presence of a feature is unrelated to the presence of other features in the same class. It is used for making probabilistic predictions based on the likelihood of features.

What is the purpose of linear regression in AI?

Linear regression in AI is used for regression modeling to discover relationships between data points and make predictions or forecasts. It works by plotting data points and finding the best fit line that represents the relationship between variables.

How does logistic regression differ from linear regression?

Logistic regression differs from linear regression in that it estimates a binary outcome (0 or 1) rather than a continuous value. It is used when the dependent variable is categorical, such as in spam filtering or predicting the occurrence of an event.

What is unsupervised learning and how does it differ from supervised learning?

Unsupervised learning is a type of AI algorithm that is given unlabeled data and creates models to find patterns or relationships within the data. It differs from supervised learning in that it does not use labeled data and instead focuses on discovering inherent structures in the data.

What is the K-means clustering algorithm and how does it work?

The K-means clustering algorithm is an unsupervised learning method that partitions data into K pre-defined clusters. It works by iteratively assigning data points to the nearest centroid and recalculating the centroids based on the assigned clusters until it converges to the best clustering.

What is the role of the Gaussian Mixture Model (GMM) in AI?

The Gaussian Mixture Model is used in unsupervised learning for clustering data into groups. It is more versatile than K-means as it allows for clusters of various shapes, not just circular, and uses a probabilistic approach rather than a distance-based one.

What is the K-Nearest Neighbors (KNN) algorithm and its applications?

The K-Nearest Neighbors (KNN) algorithm is a simple AI algorithm that classifies new data points based on their similarity to existing data points. It can be used for both supervised and unsupervised learning, with applications in classification, regression, and anomaly detection.

How do neural networks function in AI?

Neural networks function by mimicking the human brain, consisting of interconnected nodes organized into layers. They process information by adjusting connection strengths (weights) during training to recognize patterns and make predictions.

What is reinforcement learning and how does it differ from other types of learning?

Reinforcement learning is a type of AI algorithm where an agent learns by taking actions in an environment and receiving feedback in the form of rewards. It differs from other types of learning as it focuses on learning from interactions and consequences rather than from labeled data.

What are the key components of the Algorithmic Bill of Rights?

The Algorithmic Bill of Rights includes principles such as awareness, access and redress, accountability, explanation, data provenance, auditability, and validation and testing. These principles aim to guide the ethical use of algorithms and ensure fairness and transparency.

Outlines

00:00

🧠 Introduction to AI Algorithms

This paragraph introduces the concept of AI algorithms in the context of a marketing course. It explains what AI algorithms are, emphasizing their complexity and importance in programming computers to learn and operate autonomously. The paragraph outlines the necessity of training data for AI algorithms and touches on different learning patterns such as supervised, unsupervised, and reinforcement learning. It also briefly mentions the Algorithmic Bill of Rights, setting the stage for a deeper dive into various AI algorithms.

05:01

🌲 Supervised Learning Algorithms

The second paragraph delves into supervised learning algorithms, which rely on labeled data for training. It describes how these algorithms use the data to predict outcomes, likening the process to a student learning with a teacher's guidance. The paragraph highlights decision trees, random forests, support vector machines, and Naive Bayes classifiers, explaining their structures and functions. It also touches on linear regression and logistic regression, detailing how they are used for modeling relationships and estimating binary outcomes.

10:03

🔍 Unsupervised Learning and K-Means Clustering

This paragraph focuses on unsupervised learning algorithms, which work with unlabeled data to discover patterns and relationships. It introduces K-means clustering, explaining how it assigns data points to clusters based on proximity to centroids. The paragraph outlines the steps of the K-means algorithm and contrasts it with Gaussian Mixture Models, which allow for more versatile cluster shapes. It also mentions the use of unsupervised learning for applications like sales forecasting and customer churn analysis.

15:04

🤖 K-Nearest Neighbors and Neural Networks

The fourth paragraph discusses the K-nearest neighbors (KNN) algorithm, a non-parametric method used for both classification and regression. It describes how KNN classifies new data points based on proximity to existing data. The paragraph then transitions to neural networks, which are complex AI algorithms that mimic the human brain. It explains the architecture of neural networks, including input, hidden, and output layers, and how they process information to recognize patterns and make predictions.

20:04

🛠️ Reinforcement Learning and Algorithmic Rights

This paragraph explores reinforcement learning algorithms, which learn from feedback in the form of rewards. It describes the components of reinforcement learning, including the agent and the environment, and how they interact through a cycle of actions and rewards. The paragraph also covers different approaches to reinforcement learning, such as policy-based and value-based methods. It concludes with an introduction to the Algorithmic Bill of Rights, a set of principles aimed at ensuring fairness and accountability in algorithmic decision-making.

25:05

📜 Algorithmic Bill of Rights and Conclusion

The final paragraph provides a detailed look at the Algorithmic Bill of Rights, a set of guiding principles for ethical algorithm use. It discusses seven key areas: awareness, access and redress, accountability, explanation, data provenance, auditability, and validation. The paragraph emphasizes the importance of these principles in mitigating biases and ensuring transparency in algorithmic decisions. It concludes the module by summarizing the discussion on AI algorithms, their applications in various learning techniques, and the importance of adhering to the Algorithmic Bill of Rights.

Keywords

💡AI Algorithms

AI Algorithms, or Artificial Intelligence algorithms, are programming instructions that enable computers to learn and operate autonomously. They are central to the video's theme, as they form the backbone of AI applications in marketing. The script discusses various types of AI algorithms, such as supervised, unsupervised, and reinforcement learning algorithms, and their specific uses in data analysis and prediction.

💡Supervised Learning

Supervised Learning is a type of machine learning where the algorithm is trained on labeled data to predict outcomes for new data. It is a key concept in the video, as it illustrates how AI algorithms learn from existing data to make informed decisions. Examples in the script include decision trees and random forests, which are supervised learning algorithms used for classification and regression tasks.

💡Unsupervised Learning

Unsupervised Learning is another machine learning paradigm where the algorithm works with unlabeled data to identify patterns or structures within the data. The video mentions this concept in the context of algorithms like K-means clustering and Gaussian Mixture Models, which group data based on similarities without prior labeling.

💡Reinforcement Learning

Reinforcement Learning is an area of AI where an agent learns to make decisions by receiving feedback in the form of rewards or penalties. The script explains this concept by describing how the algorithm interacts with an environment, learning from the outcomes of its actions to improve performance over time.

💡Decision Trees

Decision Trees are a supervised learning algorithm represented as an inverted tree structure, where the root node represents the training dataset, and the leaf nodes represent the outcomes. The script uses decision trees as an example to explain how AI algorithms classify data by following a series of decisions based on attribute selection measures like entropy and information gain.

💡Random Forest

Random Forest is an ensemble learning method that operates by constructing multiple decision trees and outputting the class that is the mode of the classes of the individual trees. The video script describes how Random Forest improves accuracy by combining multiple decision trees, adding randomness to the model to find the best feature among a random subset of features.

💡Support Vector Machines (SVM)

Support Vector Machines (SVM) are used for classification or regression by finding the hyperplane that best separates data into classes. The script explains how SVM works by maximizing the margin between support vectors to achieve the best possible segregation of data points.

💡Naive Bayes

Naive Bayes is a classification algorithm based on Bayes' theorem that assumes the simplicity of feature independence. The video script describes Naive Bayes as a probabilistic classifier that predicts the likelihood of an event based on the presence of certain features, making probabilistic predictions despite potential overlapping attributes.

💡Linear Regression

Linear Regression is a statistical method used for predictive analysis, making it a fundamental concept in the video's discussion of AI algorithms. The script explains how linear regression models the relationship between a dependent variable and one or more independent variables, predicting continuous numeric values such as sales or age.

💡Logistic Regression

Logistic Regression is an algorithm used when the dependent variable is categorical, often taking binary values like 0 or 1. The video script illustrates its use in applications like spam filters, where it estimates the probability of an event occurring based on a logistic function that maps inputs to a probability between 0 and 1.

💡K-means Clustering

K-means Clustering is an unsupervised learning algorithm that partitions data into K distinct clusters based on the proximity of data points. The script describes the iterative process of K-means, from selecting initial centroids to assigning data points to the nearest centroid, and adjusting centroids until the clusters are optimized.

💡Gaussian Mixture Models (GMM)

Gaussian Mixture Models (GMM) are a probabilistic model that assumes the data is generated from a mixture of several Gaussian distributions. The video script contrasts GMM with K-means clustering, noting that GMM can handle more complex cluster shapes than the circular patterns of K-means, allowing for greater clarity in data segmentation.

💡K-Nearest Neighbors (KNN)

K-Nearest Neighbors (KNN) is a non-parametric algorithm used for both classification and regression. The script describes KNN as a 'lazy learning' algorithm that stores data and classifies new instances based on the nearest neighbors in the training dataset, making it useful for anomaly detection and classification tasks.

💡Neural Networks

Neural Networks are a collection of AI algorithms inspired by the human brain, consisting of interconnected nodes or neurons organized into layers. The video script explains how neural networks learn by adjusting connection strengths during training, enabling them to recognize patterns and make predictions, with applications in classification and pattern recognition.

💡Algorithmic Bill of Rights

The Algorithmic Bill of Rights is a set of guiding principles aimed at ensuring ethical use of algorithms. The video script outlines seven areas covered by these principles, including awareness, access and redress, accountability, explanation, data provenance, auditability, and validation. These principles are crucial for understanding the responsible development and application of AI algorithms.

Highlights

Introduction to AI algorithms in marketing, discussing various learning patterns: supervised, unsupervised, and reinforcement learning.

Definition of an AI algorithm as a complex set of instructions for a computer to learn and operate on its own.

Importance of training data acquisition and labeling in distinguishing different AI algorithms.

Overview of supervised learning algorithms, emphasizing their reliance on clearly labeled data for training.

Explanation of decision trees, their structure, and how they use attribute selection measures for classification.

Introduction to Random Forest, a collection of decision trees that improve accuracy through diversity.

Support Vector Machines (SVM) for classification or regression by finding the optimal hyperplane for data separation.

Naive Bayes classifier, based on the base theorem, and its use in large datasets with various classes.

Linear regression for predictive analysis and forecasting, emphasizing its simplicity and popularity.

Logistic regression for binary outcomes, using a logistic function to estimate probabilities.

Unsupervised learning algorithms and their use in creating models from unlabeled data to find patterns.

K-means clustering for dividing datasets into clusters based on data point proximity to centroids.

Gaussian Mixture Models for clustering with more versatile shapes than K-means, allowing linear patterns.

K-Nearest Neighbors (KNN) algorithm for both supervised and unsupervised learning, focusing on data proximity.

Neural networks as complex AI algorithms mimicking the human brain, with applications still being discovered.

Reinforcement learning algorithms that learn from feedback in the form of rewards, involving an agent and an environment.

Algorithmic Bill of Rights, a set of guiding principles for ethical algorithm design and use.

Seven general areas of the Algorithmic Bill of Rights, including awareness, access, accountability, and auditability.

Conclusion summarizing the module's coverage of AI algorithms, learning techniques, and ethical considerations.

Transcripts

[Music]

[Music]

welcome to this uh nptl online

certification course on artificial

intelligence in marketing and now we are

discussing module 4 so we are talking

about introduction to AI algorithms and

we are in chapter 1 and module 4 so this

is what we are talking about that is

Introduction to artificial intelligence

algorithms so to introduce the module we

will talk about what are AI algorithms

and how they

work what are the various types of

commonly used AI algorithms under the

different kind of learning patterns and

we have seen that the the various types

of learning patterns are supervised

learning unsupervised learning and

reinforcement learning and then we will

talk briefly on the algorithmic Bill of

Rights so now to start with what is an

AI algorithm the definition of algorithm

is a set of instructions to be followed

in calculation or other operations so it

is a pure simple set of

instructions this applies to both

mathematics and computer science so thus

at the essential level an AI algorithm

is the programming that tells the

computer how to learn to operate on its

own and AI algorithm is much more

complex than what most people learn

about in algebra of course a complex set

of rules Drive AI program determine

their steps and their ability to learn

without an algorithm AI would not

exist while a general algorithm can be

simple AI algorithms are by Nature more

complex AI algorithms work by taking in

training data that helps the algorithm

to learn how the data is acquired and is

labeled marks the key difference between

different types of AI algorithm so keep

in mind that how that data is acquired

and is

labeled that gives the clear key

difference between different types of AI

at the code level an AI algorithm takes

in training data labeled or unlabeled

supp lied by developers or acquired by

the program itself and use that

information to learn and grow then it

completes its task using the training

data as a basis so that training data is

so important for this kind of AI some

types of AI algorithms can be taught to

learn on their own and taken new data to

change and refine their processes others

will need the intervention of

programmers in order to streamline now

let let us look at the various types of

AI algorithms so there are three major

categories of AI algorithms that we have

already learned in the previous module

namely one was the supervised learning

the second was unsupervised learning and

the third was reinforcement learning the

key difference between these algorithms

are in how they are trained and how they

function so their differences come

from a how they are trained and B how

they function under those

categories there are dozens of different

algorithms we will discuss about the

most popular and commonly used from each

category as well as where they are

commonly used so the supervised learning

algorithms the first and the most

commonly used category of algorithm is

supervised

learning these work by taking in clearly

labeled data while being trained and

using that to learn and grow it uses the

labeled data to predict outcomes for

other data the name supervised learning

comes from the comparison of a student

learning in the presence of a teacher or

expert so that is why it is called as

supervised learning building a

supervised learning algorithm that

actually works take a team of dedicated

experts to evaluate and review the

results not to mention data scientist to

test the models the algorithms created

to ensure their accuracy against the

original data and catch any errors from

the artificial

intelligence so this is the decision

tree one of the most common supervised

learning algorithm decision trees get

their name because of their tree like

structure even though the tree is

inverted so it is the inverted tree the

roots of the tree are the training data

sets and they leads to specific

nodes which donates a text attribute

notes often lead to another notes and a

note that doesn't lead onward is called

a

leaf so this is the decision node that

is the root node then we have a sub tree

decision node Leaf node this is again

Leaf node because nothing flows from

them here in this decision node again

another decision node node and then this

is leaf node this is leaf node and this

is LEF Leaf node decision Tre is

classify all the data into decision

nodes it uses a selection criteria

called attribute selection

measures which takes into account

various measures some examples would be

entropy gain ratio Information Gain

Etc using the root data and following

the ASM the decision tree can classify

the data it is given by following the

training data into subnodes until it

reaches the conclusion a decision tree

diagram with root node decision node and

Leaf node for better understanding so

this is how root node with friends then

yes windy cold yes no below par so this

is the branch no is splitting walk or

cart decision notes walk above Park cart

cold Etc so this is the demonstration of

this decision tree diagram another type

is random Forest the random Forest

algorithm is actually a broad collection

of different recision trees leading to

its name

the random Forest builds different

decision trees and connects them to gain

more accurate results so that is the

main advantage that it gives more

accurate results this can be used for

both classification and regression type

of supervised learning while a solo

decision tree has no outcome and a

narrow range of groups the forest

assures a more accurate result with a

bigger number of groups and decisions it

has the added benefits of adding random

Ness to the model by finding the best

feature among a random subset of

features overall these benefits create a

model that has wide diversity that many

data scientist

favors so as we can see from the diagram

the results of decision Tre 1 2 and

three are combined which is then

averaged out or the majority is

considered as the final result so these

are the data sets so this is decision

Tre one this is decision tree 2C

decision tree three so with the three

typ with the same type of data sets

there are three decision trees and the

three results majority voting averaging

and then we get the final results

another is Vector support Vector

machines the support Vector machine

algorithm is another common AI algorithm

that can be used for either

classification or regression but is most

often used for

classification the support Vector

machine

works by plotting each piece of data on

a chart in N Dimension space where n is

the number of data points then the

algorithm classifies the data point by

finding the hyper plane that separates

each class there can be more than one

hyper plane the main objective of a

support Vector machine is to segregate

the given data sets in the best possible

way the distance between either nearest

points is known as the margin the the

objective is to select a hyper plane

with the maximum possible margin between

support factors in the given data set

svm searches for the maximum marginal

hyper plane so these are the two

excesses X1 and X2 and here it is a

negative hyper plane then there are

support vectors this is positive hyper

plane so that is maximum margin and this

is maximum margin hyper plane so

generate hyperplanes which segregates

the classes in the best way the figure

on the top shows three hyper

planes black blue and

orange so these are the three hyper

planes here the blue and orange have

high classification errors but the black

is separating the two classes

correctly so so this black black is

separating these two select the right

hyper plane with the maximum segregation

from the either nearest data points as

shown in figure at the bottom then there

is Nave base

the reason this algorithm is called Nave

Bas is that it is based on base theorem

and also relies heavily on a large

assumption that the presence of one

particular feature is unrelated to the

presence of other features in the same

class that major assumption is the Nave

aspect in the name so KN bu is useful

for large data set with different

classes it like many other supervised

learning algorithm is a classification

algorithm it is an algorithm that learns

the probability of every object its

features and which groups they belong to

it is also known as

probabilistic classifier for example you

cannot identify a bir based on its

feature and color as there are many

words with similar attributes but you

make a probabilistic prediction about

the same and that is where knif wise

algorithm comes in so this is these are

the three classifiers 1 2 3 and this is

knif bias classifiers

and this is how they are classified into

three different categories NWI use the

following equation that is p h upon e is

equal to p h upon e into p h ID p e so p

h upon e denotes how event H happens

when event e take

place then P eh represents how often

event e happens when event H takes place

first pH represents the probability of

event X happening on its own and P e

represents the probability of Event Y

happening on its own so then comes

linear regression is a supervised

learning AI algorithm used for

regression modeling it is mostly used

for discovering the relationship between

data points predictions and

forecasting much like support Vector

machines it works by plotting pieces of

data on a chart with the

xais as the independent variable and the

Y AIS as the dependent variable the data

points are then plotted

in a linear fashion to determine their

relationships and forecast possible

future data linear regression is one of

the easiest and the most popular machine

learning algorithm so that is the best

part that is it it is the easiest it is

a statistical method that is used for

predictive

analysis linear regressions make

predictions for continuous real or

numeric values such as the sales salary

age product prices Etc linear

integration algorithm shows a linear

relationship between a dependent that is

y and one or more independent that is X

variables hence that is called as a

linear regression since linear

regression shows the linear relationship

which means it finds how the value of

the dependent variable is changing

according to the value of the

independent variable the linear

regression model provides a sloped

straight line representing the

relationships between the variables so

this is these are the independent

variables on the x-axis and here we have

the dependent variable and then we have

all these data points and in between is

the line of regression so now it it

tells that how it will happen here what

what will the dependent variable look

look like like at this level of

independent variable then comes logistic

regression a logistic regression

algorithm usually uses a binary value 01

to estimate value from a set of

independent

variables the output of Lo logistic

regression is either 1 or zero yes or no

an example of this would be a spam

filter in email the filter uses logistic

regression to Mark whether the incoming

mail is Spam zero or not one logistic

regression is only useful when the

dependent variable is

categorical either yes or no so if it is

not if it is somewhere in between then

logistic regression will not work the

logistic regression model is based on

logistic function which is the type of s

shaped curve that Maps any continuous

input to the probability value between 0

and 1 the logistic function allows us to

model the relationship between the

independent variables and the

probability of dependent variable taking

on the value of one the logistic

regression model estimates the

coefficient of the independent variables

that are most productive of the

dependent variable these coefficients

are used to create a linear equation

that is then transformed by the logistic

functions to produce a probability value

for the dependent variable taking on the

value one the logistic regression is

commonly used in fields such as

Healthcare marketing finance and social

sciences to predict the likelihood of an

event occurring such as whether a

patient has a certain disease or or

whether the customer will buy a product

or not so now as you can see from this

figure we have independent variable X

and then we have dependent variable y

now this dependent variable varies from

0 to 1 independent variable can take any

value but dependent variable can take

values only between 0 to 1 so this

predicts y lies within 0 to one one

range

so that is why it is a s shaped

curve

so now you

see at this level what will be the value

of y at this level what will be the

value of y and at this level what will

be the value of y so as we have already

studied before unsupervised learning

algorithms are given data that is not

labeled unsupervised learning algorithm

use that unlabelled data to create

models and evaluate the relationships

between different data points in order

to give more insights to the data the

next comes K means clustering K means is

an algorithm designed to perform the

clustering

function in unsupervised learning it

does takes this by taking in the

predetermined clusters and plotting out

all the data regardless of the

cluster it then plots a randomly

selected piece of data such as the cerid

of each cluster think link of it as a

circle around each cluster which with

that piece of data as the exact center

point from there it sorts the remaining

data points into clusters based on their

proximity to each other and the CID data

point for each cluster the algorithms

takes the unlabel data sets as input

divides the data set into K numbers of

clusters and repeat the process until it

does not find the best cluster the value

of K should be predetermined in this

algorithm the means clustering algorithm

mainly perform two tasks the first is

determine the West value of K Center

points or cids by and iterative process

the second is assign each data point to

its closest G

Center those data points which are near

to the particular K Center creates a

cluster so the working of the K means

algorithm is as follows first select the

number of K to decide the number of

clusters second is Select random ke

points or or

centroids the third is assign each data

point to those closed centroid which

means form which will form the

predefined K cluster the fourth is

calculate the variance and the place a

new centroid of each cluster the fifth

is repeat the third

steps which means reassigning each data

point to the new closest Cento for each

cluster so this is how it works before

the K means and this after K means how

neatly they are clustered the next comes

gausian mixture model goian mixture

models are similar to g means clustering

in many ways both are concerned with

sorting data into predetermined clusters

based on proximity however goian models

are a little more versatile in the

shapes of the Clusters they

allow K means clustering only allows

data to be clustered in circles with the

centroid in the center of each cluster

goian mixtures can handle data that LS

on the graph in more linear patterns

allowing for oblong shaped

structures this allows for greater

Clarity in clustering of one data point

lands inside the circle of another

cluster the starting point and training

process of the K means and GMM are the

same however K means use a

distance-based approach NG GMM uses a

probabilistic based approach there is

one primary assumption in GMM the data

set consist of multiple cians in other

word a mixture of the cians it is used

to forecast the sales of product

understand customer churn through the

length of different groups of customers

some AI algorithms can use either

supervised or unsupervised data input

and still function they might have

slightly different applications based on

their

status the next comes K nearest neighbor

algorithm so K nearest neighbor that is

K NN algorithm is a simplistic AI

algorithm that assumes that all the data

points provide ided are in a proximity

to each other and plots them onto a map

to show the relationship between them

then the algorithm can calculate the

distance between data points in order to

extrapolate the relationships and

calculate the distance on the graph both

supervised and unsupervised algorithm in

supervised learning it can be used for

either classification or regression

applications in unsupervised learning it

is popularly used for anomaly detection

that is finding data that does not

belong and removing it a NN is a

non-parametric algorithm which means it

does not make any assumptions on

underlying data it is also called a lazy

learn algorithm because it does not

learn from the training set immediately

Instead at the training phase it just

store the data sets and when it gets new

data it classifies that data into a

category that is much similar to the new

data KNN algorithms can be used for

regression as well as classification but

mostly it is used for classification

problem problem s Suppose there are two

categories category a and category B and

we have a new data point X1 so this data

point will lie in which of these

categories to solve this type of problem

we need a knnn algorithm with the help

of knnn we can easily identify the

category or class of a particular data

set so now this was category a this was

category B and now we have this new data

now after KNN this new data point is

assigned to category

1 so now it becomes easier to deal with

this uh this data point rather than just

having one data point is Stand Alone the

next comes neural networks neural

network algorithm is a term for a

collection of AI algorithms that mimic

the functions of a human brain so that

mimics the functions of a human brain

these tend to be more complex than many

of the algorithms discussed above and

have applic applications which are still

being discovered so all those

applications have not yet been

discovered they are still in the process

of Discovery in unsupervised and

supervised learning it is it can be used

for

classification and pattern

recognition it consists of

interconnected nodes that is neurons

organized into layers information flows

through this

nodes and the network adjust the

connection strengths that is weights

during training to learn from data

enabling it to recognize

patterns make predictions and solve

various tasks in machine learning and

artificial intelligence and there are

three levels in the network architecture

the input layer the hidden layer which

can be more than one and the output

layer because of the numerous layers it

is sometimes refers to as the MLP that

is multi-layer percep dra it is possible

to think of the Hidden layer as a

distillation layer which ass some of the

most relevant patterns from the inputs

and send them onto the next layer for

further analysis it accelerates and

improves the efficiency of the network

by recognizing just the most important

information from the inputs and

discarding the Redundant

information so this is how it works so

this is the hidden layer then there are

input layers Network layers feed forward

so this is Network

output and this is back propagation then

comes reinforcement learning algorithm

the last major type of AI algorithm is

reinforcement learning algorithm which

learns by taking in the feedback from

the results of its action this is

typically in the form of a reward the

reinforcement algorithm is usually

composed of two major

parts the first is an agent that

performs an

action that is

one and the second is the environment in

which the action is performed so these

are the two major parts of this

algorithm the cycle begins when the

environment sends a state signal to the

agent that cues the agent to perform a

specific action within the environment

once the action is performed the

environment sends the

reward signal to the agent informing it

on what happens so that the agent can

update and evaluate its last

actions then with that new information

it can take the action again that cycle

repeats until the environment sends a

termination signal so there are two

types of reinforcement the algorithm can

use one is either a positive report or a

negative reward in reinforcement

algorithms there are slightly different

approaches depending on what is being

measured and how it is being measured

here are some definitions of different

models and measures so one is

policy the approach that agent takes to

determine the next action taken by the

agent the second is model the situation

and dynamics of the environment and the

third is value the expected long-term

results this is different from the

reward which is the result of a single

action within the environment the value

is the long-term result of many actions

the second is value based in this value

based reinforcement algorithm the agent

pushes towards an expected long-term

return so that is important

here instead of just focusing on the

short-term reward the next is policy

based a policy based reinforcement

algorithm usually take one of the two

approaches to determine the next course

of

action either uh a standardized approach

so that can be one approach where any

state produces the same action or the

dynamic approach

where certain probabilities are mapped

out and the probabili calculated each

probability has its own policy

reactions the next is the model based in

this algorithm the programmers create a

different Dynamic for each environment

so the programmer create different

Dynamics for each environment one

environment one

dynamics that way when the agent is put

into each different model

it learns to perform consistently under

each condition now let us look at the

algorithms algorithmic Bill of Rights in

January 2017 the US public policy

Council of the associations for

computing

Machinery which consist of Educators

researchers and Professionals in the

world of Information Technology outlined

a set of guiding principles that could

serve as a precursor for an algorithmic

Bill of Rights these principles covers

seven general areas which we will be

discussing one by one so the first one

is awareness those who design Implement

and use algorithms must be aware of

their potential biases and possible harm

and take these into accounts in their

practices the second is access and

redress those who are negatively

affected by algorithms must have systems

that enable them to question the

decisions and seek a red

the third is accountability

organizations that use algorithms must

take responsibility for the

decisions those algorithms reach even if

it is not feasible to explain how the

algorithms arrive at those decisions the

fourth is explanation those affected by

algorithm should be given explanations

of the

decisions and the procedures that

generated them the fifth is data

provenance those who design and use

algorithms should maintain record of the

data used to train the algorithm and

make those records available to

appropriate individuals to be studied by

to be studied for possible

biases the sixth is auditability

algorithms and data should be recorded

so that they can be audited in case of

possible harm the seventh is validation

and testing organizations that use

algorithms should test them regularly

for biases and make the results publicly

available so to conclude in this module

we have briefly introduced AI algorithms

and how do they work then we have

discussed about the commonly used AI

algorithms under the various learning

techniques that is supervised

unsupervised and reinforcement learning

and then finally we have given a brief

on the algorithmic Bill of Rights and

these are the five references from which

the material for this module was taken

thank

[Music]

you

[Music]

oh