You are on page 1of 25

The Big Question

Can We Make Computers Self Learn From Data?

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 1


What is Machine Learning?

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 2


Machine Learning is A Field of Computer Science That
Gives Computers The Ability To Learn Without Being Explicitly
Programmed
Wikipedia

Machine Learning is A Type of Artificial Intelligence (AI) That Allows


Software Applications To Become More Accurate in Predicting
Outcomes Without Being Explicitly Programmed
Tech Target

Machine Learning is The Science of Getting Computers to Act


Without Being Explicitly Programmed
Stanford University
Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 3
What is The Functional Flow of Machine Learning?

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 4


Real Time Stream The Data To
Systems Be Analyzed

Machine Apply Self-Learning Re-Analyze The


Learning Field Algorithms Data

Artificial Turn The Data into


Intelligence Field Knowledge

Is The Goal
No
Achieved?

Yes

Stop
Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 5
What Are The Different Types of Machine Learning
Approaches?

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 6


• Labeled Data

Supervised Learning • Direct Feedback

• Predict Outcome OR Future

• No Labeled

Un-Supervised Learning • No Feedback

• Find Hidden Structure in Data

• Decision Process

Reinforcement Learning • Reward Systems

• Learn Through Series of Actions


Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 7
Supervised Learning

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 8


Where Supervised Learning is More Suitable?

• Supervised Learning is More Suitable in “Making Predictions

About The Future”

What is The Main Goal of Supervised Learning?

• Learn A Model From Labeled Training Data Which Allows The


End-User To Make Predictions About Unseen OR Future Data.

What Does Supervised Learning Refers To?

• Supervised Learning Refers To A Set of Samples Where The

Desired “Output Signals Called Labels” Are Already Well


Known.
Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 9
What is The Functional Flow of Supervised Learning?

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 10


Training Data
With Labels

Apply Machine
Learning Algorithm

Inject New Develop A Finalize And Cross


Data Predictive Model Verify The Prediction

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 11


What Are The Sub-Categories of Supervised Learning?

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 12


Classification Task
• The Main Goal is To Predict The Categorical “Class Labels” of
New Operational Instances, Based on Past Observations.
• Create Possible Classes which Are Discrete To Identify By
Properties.
• Class Labels Can Be “Binary Class” OR “Multi-Class” By Nature.
• Any Label That is Not Part of The Training Dataset is Recognized
Correctly.
What is Binary Class?
• In “Binary Class” The Training Samples Are Labeled Exactly By
Two Categories.
• The Sample Gets Distributed into Two Different Dimensions on
The Graph.
What is Multi Class?
• In “Multi Class” The Training Samples Are Labeled By Multiple
Categories.
• The Sample Gets Distributed into Multi Dimensional Axis on The
Graph.
Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 13
Regression Task
• The Main Goal of “Regression Task” is To Predict For Continuous
Outcomes.
What is The Approach of Regression Analysis?
• In “Regression Analysis” The Process is Given A Number of
“Predictor Called As Explanatory Variables” And A “Continuous
Response Variable Called As Outcome OR Target”, And The
Process Should Try To “Find A Relationship Between Those
Variables That Allows The Process To Predict An Outcome”
What is The Success Factor For Regression Analysis?
• “Regression Analysis” Can Be More Successful And Accurate
When We Supply
• Huge Sample OR Training Data
• More Number of Explanatory Variables
• Proper Relationship Between The Variables
Do We Have Any Classes in Regression Analysis?
• In Practical Sense Regression Analysis Does Not Has
Classification
Wednesday, December 11, 2019of Classes.
Machine Learning By Sathish Yellanki Slide No : 14
Un-Supervised Learning

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 15


Where is Un-Supervised Learning is Applied?
• Un-Supervised Learning is Most Prominent in Discovering
Hidden Structures.
• In Un-Supervised Learning We Deal With Unlabeled Data OR
Data of Unknown Structure.
Where Un-Supervised Learning is More Suitable?
• Explore The Structure of The Data To Extract Meaningful
Information Without The Guidance of A Known “Outcome
Variable” OR A “Reward Function”.
What Does Un-Supervised Learning Refers To?
• Un-Supervised Learning Refers To A Set of Samples Where The
Desired “Output Signals Called Labels” Are Not Well Known

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 16


What is The Functional Flow of Un-Supervised Learning?

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 17


Training Data
Without Labels

Apply Machine
Learning Algorithm

Finalize And Cross


Cross Verify Develop A Clustered
Verify The
The Cluster Model
Classification

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 18


What Are The Sub-Categories of Un-Supervised Learning?

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 19


Clustering Concept OR Un-Supervised Classification
• The Main Task of Clustering Concept is To Find Sub-Groups
Within The Data.
What is Clustering Concept?
• Clustering is An Exploratory Data Analysis Technique Which
Allows The Developer To “Organize A Pile of Information into
Meaningful “Subgroups Called Clusters” Without Having
Any Prior Knowledge of Their Group Memberships”
How De We Finalize The Cluster?
• Cluster is Defined With A Group of Objects That “Share A Certain
Degree of “Similarity” But Are More “Dissimilar” To Objects”
in Other Clusters.
What is The Final Outcome of Clustering?
• Clustering is A Great Technique if Applied Properly For
“Structuring Information And Deriving Meaningful
Relationships From Data”
Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 20
Dimensionality Reduction
• The Main Task of Dimensionality Reduction is Applied For Data
Compression.
What is Dimensionality Reduction Concept?
• Un-Supervised Dimensionality Reduction is A Very Common
Technique Used To Remove Noise From Data.
What is Most Challenging in Dimensionality Reduction?
• Identifying The “Noise” Within The Analyzed Dataset.
• Analyzing The Degrading Factor Upon The Predictive
Performance of The Algorithm.
What is The Final Implementation Concept?
• Compress The Data onto A “Smaller Dimensional
Subspace” And Retain The Most of The Relevant Information.
What is The Final Outcome of Dimensionality Reduction?
• A System The is More Comprehensive And Consistent in
Presentation.
Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 21
Reinforcement Learning

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 22


Where Reinforcement Learning is Applied?
• Reinforcement Learning is Used in Solving Interactive Problems.
• In Reinforcement Learning We Interact With The Environment
To Learn The System in Continuity.
What is The Main Goal in Reinforcement Learning?
• The Goal is To Develop A “System” Which is Also Called As
“Agent” Which Can Improve its Performance Based on
“Interactions With The Environment”.
What Does Reinforcement Learning Refers To?
• In Reinforcement Learning The Information About The Current
State of The Environment Generally Includes A “Reward Signal”.
• An Agent Can Use Reinforcement Learning To Learn A Series of
Actions That Maximizes The Reward Via An Exploratory Trial-
And-Error Approach OR Deliberative Planning.
Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 23
What is The Functional Flow of Reinforcement Learning?

Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 24


Environment

Reward
State
Action

Agent

How Does Reward Maximization is Done?


• In General The Agent in Reinforcement Learning Tries To
Maximize The Reward By A Series of Interactions With The
Operational Environment.
Wednesday, December 11, 2019 Machine Learning By Sathish Yellanki Slide No : 25

You might also like