Data Science Full Course | Edureka

DATA SCIENCE CERTIFICATION TRAINING www.edureka.co/data-science
Agenda
❖ Evolution of Data
❖ Introduction To Data Science
❖ Data Science Careers and Salary
❖ Statistics for Data Science
❖ What is Machine Learning?
❖ Types of Machine Learning
❖ What is Deep Learning?

Evolution of Data
2.5 x 1018 Bytes

Evolution of Data
3 MILLION
4.3 MILLION

Data science, also known as data-driven science, is an
interdisciplinary field about scientific methods, processes, and
systems to extract knowledge or insights from data in various
forms, either structured or unstructured.
What is Data Science?

Career : Data Science
Data Analyst AI/ML Engineer Data Scientist

Data Analyst

Data Scientist

Machine Learning Engineer

Salary Trends
Average Salary (US)
Average Salary (IND)
Data Analyst Data Scientist ML Engineers

RoadMap
Earn a Bachelor’s Degree
• Computer Science
• Mathematics
• Information Technology
• Statistics
• Finance
• Economics

RoadMap
edureka!
Fine Tune Technical Skills
• Statistical methods and Packages
• R/Python and/or SAS languages
• Data warehousing and BI
• Data Cleaning, Visualization and Reporting Techniques
• Working knowledge of Hadoop & MapReduce and Machine learning techniques

RoadMap
edureka!
Develop Business Skills
• Analytic Problem-Solving
• Effective Communication
• Creative Thinking
• Industry Knowledge

RoadMap
edureka!
Master’s Degree or Certifications Programs
• MS/MTech in CS, Statistics, ML
• Big Data Certifications
• Data Analysis / Machine Learning
Certifications

RoadMap
edureka!

RoadMap
edureka!
Work on Projects
Related to Field

Data Analyst Skills
Analytical skills Communication skills Critical thinking
Attention to detail Mathematics skills Technical skills/tools

Data Scientist Skills
Analytics & Statistics Machine Learning Algorithms Problem Solving Skills
Deep Learning Business Communication Technical skills/tools

ML Engineer Skills
Programming Languages Calculus & Statistics Signal Processing
Applied Maths Neural Networks Language Processing

Statistics
Data Science Peripherals

Statistics Prog Languages

Statistics Prog Languages Software

Statistics Prog Languages Software Machine Learning

Statistics Prog Languages Software Machine Learning Big Data

What is Data ?

Variables & Research
Independent variable
Dependentvariable

Population & Sampling

Non-Probability
Probability

Non-Probability
Probability
Random sampling
Systematic sampling
Stratified sampling

Measures of Center

Measures of Spread

Skewness

Confusion Matrix
You can calculate the accuracy of your model with:

Probability
Probability is the measure of how likely something will occur

Probability
The equation describing a continuous probability distribution is called
a probability density function.

Probability
The normal distribution is a probability distribution that associates the
normal random variable X with a cumulative probability .
The normal distribution is defined by the following equation:
Y = [ 1/σ * sqrt(2π) ] * e -(x - μ)2/2σ2
Where,
X is a normal random variable.
μ is the mean and
σ is the standard deviation.

Probability
The central limit theorem states that the sampling distribution of the
mean of any independent, random variable will be normal or nearly
normal, if the sample size is large enough.

What is Machine Learning?
Machine Learning is a class of algorithms which is data-driven, i.e. unlike "normal"
algorithms it is the data that "tells" what the "good answer" is
Getting computers to program themselves and also teaching them to make decisions using data
“Where writing software is the bottleneck, let the data do the work instead.”

Features of Machine Learning

How It Works?
Learn from Data
Find Hidden Insights
Train and Grow

Applications of Machine Learning

Market Trend: Machine Learning

Machine Learning Life Cycle
Step 1 Step 2 Step 3 Step 4 Step 5 Step 6
Collecting Data
Data Wrangling
Analyse Data
Train Algorithm
Test Algorithm
Deployment

4
5
6
1
2
3
Step 1: Collecting Data

4
5
6
1
2
3
Data Wrangling

4
5
6
1
2
3
Analyse Data
model

4
5
6
1
2
3
Train Algorithm
model
Training set

4
5
6
1
2
3
Test Algorithm
model
Test set Accurate?

4
5
6
1
2
3
Operation and Optimization

Important Python Libraries

Types of Machine Learning

Supervised Learning
It’s a
Face
Label
Face
Label
Non-Face

Unsupervised Learning

Reinforcement Learning

Supervised Learning
Supervised learning is where you have input variables (x) and an output variable (Y)
and you use an algorithm to learn the mapping function from the input to the output
It is called Supervised Learning because the process of an algorithm learning from the training
dataset can be thought as a teacher supervising the learning process

Supervised Learning
Training and Testing
Prediction
Historical
Data
Random
Sampling
Training
Dataset
Testing
Dataset
Machine
Learning
Statistical
Models
Prediction
&
Testing
Model Validation
Outcome

Supervised Learning
Training and Testing
Prediction
New
Data
Model Predicted Outcome

Supervised Learning Algorithms
Linear Regression
Logistic Regression
Decision Tree
Random Forest
Naïve Bayes Classifier

Linear Regression
Linear Regression Analysis is a powerful technique used for predicting the unknown value of a
variable (Dependent Variable) from the known value of another variables (Independent Variable)
• A Dependent Variable(DV) is the variable to be predicted or explained in
a regression model
• An Independent Variable(IDV) is the variable related to the dependent
variable in a regression equation

Simple Linear Regression
Dependent Variable Independent Variable
Y = a + bX
Y - Intercept Slope of the Line

Regression Line
Linear Regression Analysis is a powerful technique used for predicting the unknown value of a variable
(Dependent Variable) from The regression line is simply a single line that best fits the data
(In terms of having the smallest overall distance from the line to the points)
Fitted Points
Regression Line

Copyright © 2019, edureka and/or its affiliates. All rights reserved.
Demo

Hi I am john, I need some
baseline for pricing my Villas
and Independent Houses
Real Estate Company Use Case

Dataset Description
Column Description
CRIM per capita crime rate by town
ZN proportion of residential land zoned for lots over 25,000 sq.ft.
INDUS proportion of non-retail business acres per town.
CHAS Charles River dummy variable (1 if tract bounds river; 0 otherwise)
NOX nitric oxides concentration (parts per 10 million)
RM average number of rooms per dwelling
AGE proportion of owner-occupied units built prior to 1940
DIS weighted distances to five Boston employment centres
RAD index of accessibility to radial highways
TAX full-value property-tax rate per $10,000
PTRATIO pupil-teacher ratio by town
B 1000(Bk - 0.63)^2 where Bk is the proportion of blacks by town
LSTAT % lower status of the population
MEDV Median value of owner-occupied homes in $1000's

Steps
Step 1 Step 2 Step 3 Step 4 Step 5 Step 6
Collecting Data
Data Wrangling
Analyse Data
Train Algorithm
Test Algorithm
Deployment

Model Fitting
Fitting a model means that you're making your algorithm learn the relationship between predictors
and outcome so that you can predict the future values of the outcome .
So the best fitted model has a specific set of parameters which best defines the problem at hand

Types of Fitting
Machine Learning algorithms first attempt to solve the problem of under-fitting; that is, of taking a line
that does not approximate the data well, and making it to approximate the data better.

Need For Logistic Regression
WHO WILL WIN ?
Here, the best fit line in linear regression
is going below 0 and above 1

What is Logistic Regression?
The outcome(result)
will be binary(0/1)
0- If malignant
1- If benign
Logistic Regression is a statistical method for analysing a dataset in which there are one or more
independent variables that determine an outcome.
Outcome is a binary class type.

What is Logistic Regression?
Based on the threshold value set, we
decide the output from the function
The Logistic Regression Curve is called as “Sigmoid Curve”, also known as S-Curve

What is Polynomial Regression?
When we have non linear data, which can’t be predicted with a linear model. We switch to
Polynomial Regression.
Such a scenario is shown in the below graph

What is a Decision Tree?
A decision tree is a tree-like structure in which internal node represents test on an attribute
• Each branch represents outcome of test and each leaf
node represents class label (decision taken after
computing all attributes)
• A path from root to leaf represents classification rules.
True False

Building a Decision Tree

What is Random Forest?
Random Forest is an ensemble classifier made using many Decision tree models
What are Ensemble models? How is it better from Decision Trees ?

What is Naïve Bayes?
Random Forest is an ensemble classifier made using many Decision tree models
What are Ensemble models? How is it better from Random Forest?

What is Naïve Bayes ?
Naive Bayes is a simple but surprisingly powerful algorithm
for predictive modeling.
Bayes

Bayes’ Theorem
Given a hypothesis H and evidence E , Bayes' theorem states
that the relationship between the probability of the hypothesis
before getting the evidence P(H) and the probability of the
hypothesis after getting the evidence P(H|E) is
P(H|E) = P(E|H).P(H)
P(E)

Bayes’ Theorem Example

P(King) = 4/52 =1/13

P(King) = 4/52 =1/13
P(King|Face) = P(Face|King).P(King)
P(Face)

P(King) = 4/52 =1/13
P(Face)
P(Face|King) = 1

P(King) = 4/52 =1/13
P(Face)
P(Face|King) = 1
P(Face) =12/52 = 3/13

Bayes’ Theorem Proof
P(H|E) = P(E|H).P(H)
P(E)
Likelihood
How probable is the evidence
Given that our hypothesis is true?
Posterior
How probable is our Hypothesis
Given the observed evidence?
(Not directly computable)
Prior
How probable was our hypothesis
Before observing the evidence?
Marginal
How probable is the new evidence
Under all possible hypothesis?

Naïve Bayes: Working
PYTHON CERTIFICATION TRAINING www.edureka.co/python

Classification Steps

Industrial Use Cases
News Categorization Spam Filtering
Weather Predictions

Unsupervised Learning
Unsupervised learning is a type of machine learning algorithm used to draw inferences from datasets
consisting of input data Without labelled responses

Unsupervised Learning: Process Flow
Training data is collection of
information without any label

What is Clustering?
“Clustering is the process of dividing the datasets into
groups, consisting of similar data-points”
It means grouping of objects based on the information
found in the data, describing the objects or their
relationship

Why is Clustering Used?
The goal of clustering is to determine the intrinsic grouping
in a set of Unlabelled Data
Sometimes, Partitioning is the goal

Where is it used?

Types of Clustering
Exclusive Clustering
Overlapping Clustering
Hierarchical Clustering
K-Means Clustering

Types of Clustering
C-Means Clustering

Types of Clustering

K-Means Clustering
The process by which objects are classified into a
predefined number of groups so that they are as
much dissimilar as possible from one group to
another group, but as much similar as possible
within each group.

K-Means Algorithm Working

First we need to decide the number of clusters to be made. (Guessing)
K-Means Clustering : Steps
1
Let’s assume , Number of clusters = 3

Then we provide centroids of all the clusters. (Guessing)
1
2

The Algorithm calculates Euclidian distance of the points from each
centroid and assigns the point to the closest cluster.
1
2
3

Next the Centroids are calculated again, when we have our new cluster.
1
2
3
4

The distance of the points from the centre of clusters are calculated
again and points are assigned to the closest cluster.
1
2
3
4
5

And then again the new centroid for the cluster is calculated.
1
2
3
4
5
6

And then again the new centroid for the cluster is calculated.
These steps are repeated until we have a repetition in centroids or new
centroids are very close to the previous ones.
1
2
3
4
5
6
7

How to Decide the number of Clusters
The Elbow Method :
First of all, compute the sum of squared error (SSE) for some values of k (for
example 2, 4, 6, 8, etc.). The SSE is defined as the sum of the squared distance
between each member of the cluster and its centroid. Mathematically:

Pros and Cons: K-Means Clustering
• Simple, understandable
• Items automatically assigned to clusters
• Must define number of clusters
• All items forced into clusters
• Unable to handle noisy data and outliers

Fuzzy C – Means Clustering
Fuzzy C-Means is an extension of K-Means, the popular simple clustering technique
Fuzzy clustering (also referred to as soft clustering) is a form of Clustering
in which each data point can belong to more than one cluster

Pros and Cons: C-Means Clustering
• Allows a data point to be in multiple clusters
• A more natural representation of the behaviour of genes
• Genes usually are involved in multiple functions
• Need to define c, the number of clusters
• Need to determine membership cut-off value
• Clusters are sensitive to initial assignment of centroids
• Fuzzy c-means is not a deterministic algorithm

Hierarchical clustering is an alternative approach which builds a hierarchy from the bottom-up,
and doesn’t require us to specify the number of clusters beforehand

Pros and Cons: Hierarchical Clustering
• No assumption of a particular number of clusters
• May corresponds to meaningful taxonomies
• Once a decision is made to combine two clusters, it can’t be
undone
• Too slow for large datasets

Market Basket Analysis
Market basket analysis explains the combinations of products that frequently
co-occur in transactions.
Market Basket Analysis algorithms
1. Association Rule Mining
2. Apriori

Association Rule Mining
Association rule mining is a technique that shows how items are associated to each other.
Customer who purchase laptops are more likely to purchase laptop bags.
Customer who purchase bread have a 60% likelihood of also purchasing jam.
Example:

A B
Example of Association rule
➢ It means that if a person buys item A then he will also buy item B
➢ Three common ways to measure association:
Support
Confidence
Lift

Support gives fraction of
transactions which contains
the item A and B
𝑓𝑟𝑒𝑞 𝐴, 𝐵
𝑁
Support =
Confidence gives how often
the items A & B occur
together, given no. of times
A occurs
𝑓𝑟𝑒𝑞 𝐴, 𝐵
𝑓𝑟𝑒𝑞(𝐴)
Confidence =
Lift indicates the strength of
a rule over the random co-
occurrence of A and B
𝑆𝑢𝑝𝑝𝑜𝑟𝑡
𝑆𝑢𝑝𝑝 𝐴 𝑥 𝑆𝑢𝑝𝑝(𝐵)
Lift =

Association Rule Mining Example
T1 A B C
T2 A C D
T3 B C D
T4 A D E
T5 B C E
A B C A C D B C D
A D E B C E
Transactions at a local store
Set of items {A, B, C, D, E}
Set of transactions {T1, T2, T3, T4, T5}

Association Rule Mining Example
Consider the following
association rules:
1. A D
2. C A
3. A C
4. B & C A
=>
=>
=>
=>
Rule Support Confidence Lift
A D 2/5 2/3 10/9
C A 2/5 2/4 5/6
A C 2/5 2/3 5/6
B, C A 1/5 1/3 5/9
=>
=>
=>
=>
Calculate support, confidence and lift for these rules:

Apriori Algorithm
Apriori algorithm uses frequent item sets to generate association rules. It is based on the concept
that a subset of a frequent itemset must also be a frequent itemset.
But what is a
frequent item set?
Frequent Itemset: an itemset
whose support value is greater
than a threshold value.
Example: If {A,B} is a frequent
item set, then {A} and {B} should
be frequent item sets

Apriori Algorithm
Consider the following transactions:
TID Items
T1 1 3 4
T2 2 3 5
T3 1 2 3 5
T4 2 5
T5 1 3 5
Min. support count = 2
Note:
Now the first step is to
build a list of item sets of
size one by using this
transactional dataset.

Apriori Algorithm – First Iteration
TID Items
T1 1 3 4
T2 2 3 5
T3 1 2 3 5
T4 2 5
T5 1 3 5
Step 1: Create item sets of size one & calculate their support values
Itemset Support
{1} 3
{2} 3
{3} 4
{4} 1
{5} 4
Itemset Support
{1} 3
{2} 3
{3} 4
{5} 4
Table: C1|
Table: F1|
Item sets with support value less than min. support value (i.e. 2) are eliminated

Apriori Algorithm – Second Iteration
TID Items
T1 1 3 4
T2 2 3 5
T3 1 2 3 5
T4 2 5
T5 1 3 5
Step 2: Create item sets of size two & calculate their support values.
All the combinations of item sets in F1| are used in this iteration
Itemset Support
{1,2} 1
{1,3} 3
{1,5} 2
{2,3} 2
{2,5} 3
{3,5} 3
Itemset Support
{1,3} 3
{1,5} 2
{2,3} 2
{2,5} 3
{3,5} 3
Table: C2|
Table: F2|
Item sets with support less than 2 it are eliminated

Apriori Algorithm – Third Iteration
TID Items
T1 1 3 4
T2 2 3 5
T3 1 2 3 5
T4 2 5
T5 1 3 5
Step 3: Create item sets of size three & calculate their support values.
All the combinations of item sets in F2| are used in this iteration
Itemset Support
{1,2,3}
{1,2,5}
{1,3,5}
{2,3,5}
Table: C3|
Before calculating support
values, let’s perform
pruning on the dataset!

Apriori Algorithm – Pruning
TID Items
T1 1 3 4
T2 2 3 5
T3 1 2 3 5
T4 2 5
T5 1 3 5
After the combinations are made, divide C3| item sets to check if there
any other subsets whose support is less than min support value.
Itemset Support
{1,3} 3
{1,5} 2
{2,3} 2
{2,5} 3
{3,5} 3
Table: C3|
Table: F2|
If any of the subsets of these item sets are not there in FI2 then we remove that itemset
Itemset In F2|?
{1,2,3}
{1,2},{1,3},{2,3}
NO
{1,2,5}
{1,2},{1,5},{2,5}
NO
{1,3,5}
{1,5},{1,3},{3,5}
YES
{2,3,5}
{2,3},{2,5},{3,5}
YES

Apriori Algorithm – Fourth Iteration
TID Items
T1 1 3 4
T2 2 3 5
T3 1 2 3 5
T4 2 5
T5 1 3 5
Using the item sets of C3|, create new itemset C4|.
Itemset Support
{1,2,3,5} 1
Table: F3|
Table: C4|
Since support of C4| is less than 2, stop and return to the previous itemset, i.e. CI3
Itemset Support
{1,3,5} 2
{2,3,5} 2

Apriori Algorithm – Subset Creation
Frequent Item set F3|
Itemset Support
{1,3,5} 2
{2,3,5} 2
• Generate all non empty subsets for each frequent item set
❖ For I = {1,3,5}, subsets are {1,3}, {1,5}, {3,5}, {1}, {3}, {5}
❖ For I = {2,3,5}, subsets are {2,3}, {2,5}, {3,5}, {2}, {3}, {5}
• For every subsets S of I, output the rule:
S → (I-S) (S recommends I-S)
if support(I)/support(S) >= min_conf value
Let’s assume our
minimum confidence
value is 60%

Apriori Algorithm – Applying Rules
Applying rules to item sets of F3|:
1. {1,3,5}
✓ Rule 1: {1,3} → ({1,3,5} - {1,3}) means 1 & 3 → 5
Confidence = support(1,3,5)/support(1,3) = 2/3 = 66.66% > 60%
Rule 1 is selected
✓ Rule 2: {1,5} → ({1,3,5} - {1,5}) means 1 & 5 → 3
Confidence = support(1,3,5)/support(1,5) = 2/2 = 100% > 60%
Rule 2 is selected
✓ Rule 3: {3,5} → ({1,3,5} - {3,5}) means 3 & 5 → 1
Confidence = support(1,3,5)/support(3,5) = 2/3 = 66.66% > 60%
Rule 3 is selected
TID Items
T1 1 3 4
T2 2 3 5
T3 1 2 3 5
T4 2 5
T5 1 3 5

Apriori Algorithm – Applying Rules
Applying rules to item sets of F3|:
1. {1,3,5}
✓ Rule 4: {1} → ({1,3,5} - {1}) means 1 → 3 & 5
Confidence = support(1,3,5)/support(1) = 2/3 = 66.66% > 60%
Rule 4 is selected
✓ Rule 5: {3} → ({1,3,5} - {3}) means 3 → 1 & 5
Confidence = support(1,3,5)/support(3) = 2/4 = 50% <60%
Rule 5 is rejected
✓ Rule 6: {5} → ({1,3,5} - {5}) means 5 → 1 & 3
Confidence = support(1,3,5)/support(3) = 2/4 = 50% < 60%
Rule 6 is rejected
TID Items
T1 1 3 4
T2 2 3 5
T3 1 2 3 5
T4 2 5
T5 1 3 5

What is Reinforcement Learning?
Reinforcement learning is a type of Machine Learning where an agent learns to
behave in an environment by performing actions and seeing the results

Analogy
Scenario 1: Baby starts crawling and makes it to the candy

Analogy
Scenario 2: Baby starts crawling but falls due to some hurdle in between

Analogy

Reinforcement Learning Definitions

Agent: The RL algorithm that learns from trial and error
Environment: The world through which the agent moves
Action (A): All the possible steps that the agent can take
State (S): Current condition returned by the environment

Reward (R): An instant return from the environment to appraise the last action
Policy (π): The approach that the agent uses to determine the next action based on
the current state
Value (V): The expected long-term return with discount, as opposed to the short-term
reward R
Action-value (Q): This similar to Value, except, it takes an extra parameter, the current
action (A)

Reward Maximization
Reward maximization theory states that, a RL agent must be trained in such a way that, he
takes the best action so that the reward is maximum.
Agent
Opponent
Reward

Exploration & Exploitation
Agent
Opponent
Reward
Exploitation is about using the already known exploited information to heighten the rewards
Exploration is about exploring and capturing more information about an environment

The K-Armed Bandit Problem
• The K-armed bandit is a metaphor representing a casino slot machine with k pull levers (or arms)
• The user or customer pulls any one of the levers to win a predefined reward
• The objective is to select the lever that will provide the user with the highest reward

The Epsilon Greedy Algorithm
Takes whatever action seems best at the present moment

The Epsilon Greedy Algorithm
With probability 1 – epsilon, the Epsilon-Greedy
algorithm exploits the best known option
With probability epsilon / 2, the Epsilon-Greedy algorithm explores
the best known option
With probability epsilon / 2, the Epsilon-Greedy algorithm explores
the worst known option

Markov Decision Process
The mathematical approach for mapping a solution in reinforcement learning is called
Markov Decision Process (MDP)
The following parameters are used to attain a solution:

Markov Decision Process- Shortest Path Problem
A
B
C
D
30
-20
-10
10
50
15
Goal: Find the shortest path between A and D with minimum possible cost
In this problem,
• Set of states are denoted by nodes i.e. {A, B, C, D}
• Action is to traverse from one node to another {A -> B, C -> D}
• Reward is the cost represented by each edge
• Policy is the path taken to reach the destination {A -> C -> D}

Understanding Q-Learning With An Example
• 5 rooms in a building connected by
doors
• each room is numbered 0 through 4
• The outside of the building can be
thought of as one big room (5)
• Doors 1 and 4 lead into the building
from room 5 (outside)
Place an agent in any one of the rooms (0,1,2,3,4) and the goal is to reach outside the building (room 5)
0
4
3
1
2
5

0
4
3
1
2
5
Let’s represent the rooms on a graph, each room as a node, and each door as a link
1
2 3
40
5 Goal

Next step is to associate a reward value to each door:
• doors that lead directly to the goal have a
reward of 100
• Doors not directly connected to the target
room have zero reward
• Because doors are two-way, two arrows are
assigned to each room
• Each arrow contains an instant reward value
1
2 3
40
5 Goal
00
0
0
0
0
0 0
0
0
100
100
100

The terminology in Q-Learning includes the terms state and action:
• Room (including room 5) represents a state
• agent's movement from one room to another represents an action
• In the figure, a state is depicted as a node, while "action" is represented by the arrows
Example (Agent traverse from room 2 to room5):
1. Initial state = state 2
2. State 2 -> state 3
3. State 3 -> state (2, 1, 4)
4. State 4 -> state 5
1
2 3
40
5 Goal
00
0
0
0
0
0 0
0
0
100
100
100

We can put the state diagram and the instant reward values into a reward table, matrix R.
The -1's in the table represent null values
1
2 3
40
5 Goal
00
0
0
0
0
0 0
0
0
100
100
100
0 1 2 3 4 5
-1 -1 -1 -1 0 -1
-1 -1 -1 0 -1 100
-1 -1 -1 0 -1 -1
-1 0 0 -1 0 -1
0 -1 -1 0 -1 100
-1 0 -1 -1 0 100
0
1
2
3
4
5
State
Action
R =

Add another matrix Q, representing the memory of what the agent has learned through experience.
• The rows of matrix Q represent the current state of the agent
• columns represent the possible actions leading to the next state
• Formula to calculate the Q matrix:
Q(state, action) = R(state, action) + Gamma * Max [Q(next state, all actions)]
The Gamma parameter has a range of 0 to 1 (0 <= Gamma > 1).
• If Gamma is closer to zero, the agent will tend to consider only
immediate rewards.
• If Gamma is closer to one, the agent will consider future rewards
with greater weight
Note

Q – Learning Algorithm
1
2
3
4
5
6
7
8
9
Set the gamma parameter, and environment rewards in matrix R
Initialize matrix Q to zero
Select a random initial state
Set initial state = current state
Select one among all possible actions for the current state
Using this possible action, consider going to the next state
Get maximum Q value for this next state based on all possible actions
Compute: Q(state, action) = R(state, action) + Gamma * Max[Q(next state, all actions)]
Repeat above steps until current state = goal state

Q – Learning Example
First step is to set the value of the learning parameter Gamma = 0.8, and the initial state as Room 1.
Next, initialize matrix Q as a zero matrix:
• From room 1 you can either go to room 3 or 5, let’s select room 5.
• From room 5, calculate maximum Q value for this next state based on all possible actions:
Q(state, action) = R(state, action) + Gamma * Max[Q(next state, all actions)]
Q(1,5) = R(1,5) + 0.8 * Max[Q(5,1), Q(5,4), Q(5,5)] = 100 + 0.8 * 0 = 100
1
3
5
1
4
5

First step is to set the value of the learning parameter Gamma = 0.8, and the initial state as Room 1.
Next, initialize matrix Q as a zero matrix:
• From room 5, calculate maximum Q value for this next state based on all possible actions
Q(1,5) = R(1,5) + 0.8 * Max[Q(5,1), Q(5,4), Q(5,5)] = 100 + 0.8 * 0 = 100
1
3
5
1
4
5

For the next episode, we start with a randomly chosen initial state, i.e. state 3
• From room 3 you can either go to room 1,2 or 4, let’s select room 1.
Q(3,1) = R(3,1) + 0.8 * Max[Q(1,3), Q(1,5)]= 0 + 0.8 * [0, 100] = 80
The matrix Q get’s updated
3
1
2
4
3
5

For the next episode, the next state, 1, now becomes the current state. We repeat the inner loop of the Q
learning algorithm because state 1 is not the goal state.
Q(1,5) = R(1,5) + 0.8 * Max[Q(5,1), Q(5,4), Q(5,5)] = 100 + 0.8 * 0 = 100
The matrix Q remains the same since, Q(1,5) is already fed to the agent
1
3
5
1
4
5

AI and ML and DL

Limitations of Machine Learning
Traditional Machine Learning algorithms have failed to solve crucial problems of AI, such as
Natural language processing, image recognition and so on.

Limitations of Machine Learning

Deep Learning to Rescue

What is Deep Learning ?
Deep Learning is a subset of Machine Learning where similar Machine Learning Algorithms are
used to train Deep Neural Networks so as to achieve better accuracy in those cases where the
former was not performing up to the mark.
Basically, Deep learning mimics the way our brain functions i.e. it learns from experience.

Applications of Deep Learning
• Automatic Machine Translation
• Object Classification in Photographs
• Automatic Handwriting Generation
• Character Text Generation
• Image Caption Generation
• Colorization of Black and White Images
• Automatic Game Playing

Applications of Deep Learning

How Neuron Works?
Deep learning is a form of machine learning that uses a model of computing that's very much
inspired by the structure of the brain, so lets understand that first.

A Perceptron
Each neuron has a set of inputs, each of which is given a specific weight. The neuron
computes some function on these weighted inputs and gives the output.

Role of Weights and Bias
• For a perceptron, there can be one more input called bias
• While the weights determine the slope of the classifier line, bias allows us to shift the line towards left or right

Activation Functions
1.Linear or Identity
2.Unit or Binary Step
3.Sigmoid or Logistic
4.Tanh
5.ReLU
6.SoftMax

LINEAR UNIT STEP

Sigmoid Tan h

ReLu SoftMax

Perceptron Example

What are Tensors?
Tensors are the standard way of representing data
in deep learning
Tensors are just multidimensional arrays, an extension of 2-dimensional
tables (matrices) to data with higher dimension.
Tensor 0f Dimension -6 Tensor 0f Dimension –[6,4] Tensor 0f Dimension –[6,4,2]

What is TensorFlow?
In Tensorflow, computation is approached as a
dataflow graph
Tensor Flow
ADD
MATMUL
Data
RESULT

Perceptron Problems
• Single-Layer Perceptrons cannot classify non-linearly separable data points.
• Complex problems, that involve a lot of parameters cannot be solved by Single-Layer Perceptrons.

Deep Neural Network

Training Network Weights
• We can estimate the weight values for our training data using ‘stochastic gradient descent’ optimizer.
• Stochastic gradient descent requires two parameters:
• Learning Rate: Used to limit the amount each weight is corrected each time it is updated.
• Epochs: The number of times to run through the training data while updating the weight.
• These, along with the training data will be the arguments to the function.

MNIST Dataset

Companies Hiring

Data Science Master Program

YouTube Video Link in the Description

Data Science Full Course | Edureka

More Related Content

What's hot (20)

Similar to Data Science Full Course | Edureka (20)

More from Edureka! (20)

Recently uploaded (20)

Data Science Full Course | Edureka