SlideShare a Scribd company logo
Contribute
Significantly to the
Field of AI
Sambit Chakraborty
AI Solution Architect, TrainingMug
Narula Institute of Technology
• Introduction to AI Domain
• Relation with Data Science
• Different Learning Methods and familiarisation with Neural Networks
• Hello TENSOR 👋
• Get cleared about your ROLE
• AI, AI, AI…..DON’T SPAM ❌
• The third CONTRACT 🔑
• How important the DATASET is ??? 📊
Agenda
Intro To AI
Narula Institute of Technology
Let’s look at the difference b/w them
AI (Artificial Intelligence) is a program which mimics
the human intelligence, sense, reasoning, acting and
adapting.
AI programs can be divided into two types:
1. Narrow AI
2. General AI
AI vs ML vs DL
ML (Machine Learning) is subfield of AI, which
actually uses statistical algorithms to establish a
pattern & relationships within the dataset.
● The model and algorithm improves themselves day
by day by performing tasks and learning from the
experience.
● Neural Network layer count is < 3 & Node count is
< 4
DL (Deep Learning) is subset of ML, which is comprised
of more advanced algorithms, which permits
the model to train itself from dynamic datasets by
exposing multilayered NNs on experience basis, to
perform more complex tasks like speech recognition, text
completion, multimodal action pipelines etc.
● It learns the hierarchical representation of data.
● Neural Network layer count is > 3 & Node count is > 4
Relation with Data
Science
Narula Institute of Technology
At first, let’s take a look at Big Data & Data Science:
Big Data: It’s a blanket term for any collection of data sets so large or
complex that it becomes very much difficult to process them using
traditional data management technologies like RDBMS. This widely
adopted has long been regarded as one-size-fits-all solution, but the Big
Data handling have shown the otherwise demands.
Data Science: It involves using statistical and knowledge representation
methods to analyze massive amounts of complex & variety of data and
extract the knowledge it contains.
Those extracted datas and knowledges are used to train the models in
ML & DL.
What I’ll be expecting from you?
- You can think the relationship between crude oil and an oil refinery.
Venn Diagram of AI, ML, DL & Data Science
Artificial Intelligence
Machine Learning
Data Science
Deep
Learning
Learning Methods
& Neural Networks
Narula Institute of Technology
1. Supervised Learning: In this method, the entire learning of the machine is
based on labeled data.
- For each input, we know the desired output
- Algorithm is sufficient enough to map the input to the particular
correct output, which involves the minimisation process of the
difference b/w predicted output and true output.
Examples: Linear Regression, Logistic Regression, SVM, Decision Trees
etc.
Different Learning Methods
2. Un-Supervised Learning: In this method, the entire learning of the
machine is based on unlabeled data.
OKK…What does it means then?
- For each input, we don’t have any desired output.
- Goal: To discover patterns / structures within the dataset
- For this, the algorithms must be curated very carefully and
precisely to achieve this goal.
Examples: PCA (Principal Component Analysis) Method, Clustering
Methods,
Auto-Encoder Injection Method etc
3. Reinforcement Learning: In this method, the training of the model is based on
feedback & trial-error method, from its surroundings.
Ewww….what’s this now???
- Basically an agent receives information about its environment and learns
to choose actions that will maximize some reward. Like, suppose you’re
building a NN that “looks” at a video game screen and outputs game
actions that will gain positive scores, can be trained via this Learning
Method.
Examples: Different Methods including Deep Reinforcement Learning, Q-
Learning
4. Self-Supervised Learning: This is a specific instance of Supervised Learning,
but it’s different enough to have its own category.
Now it’s your turn to say the differences…..
- This Supervised Learning process doesn’t involves humans.
- The labels are automatically generated from input datas, typically using a
heuristic algorithm.
Examples: Auto-Encoders
** Self-Supervised Learning can be reinterpreted as either supervised or
unsupervised learning, depending on whether you pay attention to the learning
mechanism or to the context of its application.
Familiarisation with Neural Networks
Neural Network is itself an inbuilt algorithm which is improvised from the
Human Brain Structure and Cognitions.
- It consists of multiple layer of Interconnected artificial neurons known as
nodes or vertices. They work together to learn from data and proceed
further.
- Each of the nodes in one particular layer receives one input signal from
previous layer nodes and hence applies an activation function (like Sigmoid,
ReLU etc.) to produce the further output.
- Each of the nodes of each layers are connected with weights.
- Weights must be adjusted during the training process, otherwise the model
will not work as per your desire.
Types of NNs: FNN (Feedforward), RNN (Recurrent), CNN (Convolutional), DNN
(Deep)
How the Networks are built:
1. Input Nodes: It receives data input and then passes them to the next layer of
nodes (Hidden Layer). And has no activation function.
2. Hidden Nodes: This is the main functional block of a ML/DL model which
actually performs the computations, knowledge extraction, decisions on the
input data coming from input layer. All the nodes have ReLU activation
function (mostly) to determine the output and target node from successive
layer.
3. Output Nodes: These nodes are the final layer of a NN, which produces the
final output. It’s design is completely based on the performant Hidden Layers,
and have activation functions
- Suppose, for binary classification, we’re using single output node, for that
we’ll use Sigmoid Function
- For multiclass classification, there are multiple output nodes, for that
we’ll use Softmax Function
Hello TENSOR 👋
Narula Institute of Technology
Meet our new member: TENSOR
It’s nothing than a Data Structure
At its core, a Tensor is a container for data — almost always numerical data.
So, it’s a container for numbers.
You’re already familiar with matrices, which are nothing but 2D Tensors: as
they are a generalisation of matrices to an arbitrary number of dimensions.
** Dimension is often called an axis in the context of Tensors.
Types of Tensors
1. Scalars (0D tensors): A tensor that contains only one number is called a scalar
(or scalar tensor, or 0-dimensional tensor, or 0D tensor). The number of axes of
a tensor is also called its rank. A scalar tensor has 0 axes
Notation: array(12)
1. Vectors (1D tensors): An array of numbers is called a vector, or 1D tensor. A 1D
tensor is said to have exactly one axis.
Notation: array([12, 3, 6, 14])
This vector has five entries and so is called a 5-dimensional vector. Don’t confuse a 5D vector with a 5D tensor! A 5D vector has
only one axis and has five dimensions along its axis, whereas a 5D tensor has five axes (and may have any number of
dimensions along each axis). Dimensionality can denote either the number of entries along a specific axis (as in the case of our
5D vector) or the number of axes in a tensor (such as a 5D tensor), which can be confusing at times.
3. Matrices (2D tensors): An array of vectors is a matrix, or 2D tensor. A matrix
has two axes (often referred to rows and columns). You can visually interpret a
matrix as a rectangular grid of numbers.
Notation: array([[5, 78, 2, 34, 0],
[6, 79, 3, 35, 1],
[7, 80, 4, 36, 2]])
3. 3D tensors and higher-dimensional tensors: If you pack such matrices in a
new array, you obtain a 3D tensor, which you can visually interpret as a cube of
numbers.
Notation: array([[[5, 78, 2, 34, 0], [6, 79, 3, 35, 1], [7, 80, 4, 36, 2]],
[[5, 78, 2, 34, 0], [6, 79, 3, 35, 1], [7, 80, 4, 36, 2]],
[[5, 78, 2, 34, 0], [6, 79, 3, 35, 1], [7, 80, 4, 36, 2]]])
By packing 3D tensors in an array, you can create a 4D tensor, and so on. In deep learning, you’ll generally
manipulate tensors that are 0D to 4D, although you may go up to 5D if you process video data.
Re-imagine
what’s possible
with Generative
AI
FirstName LastName
Title, Company
Narula Institute of technology
I explained Gen AI to my
fellow mates telling them it's
like Wikipedia on the exact
thing you want to know" 🚀
Narula Institute of technology
RAJ SAHA
Google DSC’23 Lead
What is Generative AI?
Do I even need to tell?
SCAN IT
When you listen the
term
“Generative AI”
What is Generative AI?
Generative AI is a subset of artificial intelligence that uses machine learning
models to generate data indistinguishable from real, human-generated data.
It enables machines to creatively build content such as images, text, and
sound.
1. Can generate human-like content
2. Enables machines to create content that has not previously existed
3. Applications in diverse fields
What is possible
with Gen AI?
What is the Product of
Generative AI?
A revolutionary Al model capable of:
● Writing different creative text formats:
● Translating languages
● Answering questions in an informative
way
● Understanding information from text,
images, audio, and code
The Trend of Generative AI
How it started:
The Trend of Gen AI
How it may go:
● The market size in the Generative AI
market is projected to reach ₹5.41tn in
2024.
● The market size is expected to show an
annual growth rate (CAGR 2024-2030) of
22.24%, resulting in a market volume of
₹18.05tn by 2030.
● In global comparison, the largest market
size will be in the United States
(₹1,883.00bn in 2024).
Source:
73%
Of the Indian population surveyed
are using generative AI.
67%
IT leaders surveyed
said they have prioritized generative AI for their
business within the next 18 months
33%
IT leaders surveyed
said it was a top priority
Beyond the Trend
● Using LLMs to generate mock interview questions and responses
○ Example: https://g.co/bard/share/9342595e1896
● Homework grading
○ Example: https://g.co/bard/share/9681cb391ce4
● Language Trainer
○ Example: https://g.co/bard/share/901087ae1fe8
Gen AI Studio
on Vertex AI
GenAI on Google Cloud Platform
● On Vertex AI, Gen AI is available as a part of the platform with the name:
Generative AI Studio. This is the most flexible offering.
GenAI on Google Cloud Platform
● On Vertex AI, Gen AI is available as a part of the platform with the name:
Generative AI Studio. This is the most flexible offering.
● Gen App Builder (now, Vertex AI Search and Conversation) allows developers
to quickly build APIs that provide search or chat functionalities, powered by
LLMs. This is a much more niche product and cuts development time greatly.
GenAI on Google Cloud Platform
● On Vertex AI, Gen AI is available as a part of the platform with the name:
Generative AI Studio. This is the most flexible offering.
● Gen App Builder (now, Vertex AI Search and Conversation) allows developers to
quickly build APIs that provide search or chat functionalities, powered by LLMs.
This is a much more niche product and cuts development time greatly.
● Google AI Studio (MakerSuite): a freely available developer tool that
provides access to PaLM2 API along with an interface for rapidly testing
and developing LLM prompts.
Thinking AI-First
In Production
Increasingly, more companies are putting up products online that use generative
AI. While the results are often delightful, there are several challenges
In Production
Increasingly, more companies are putting up products online that use generative
AI. While the results are often delightful, there are several challenges
● Cost of putting generative AI applications in production
In Production
Increasingly, more companies are putting up products online that use generative
AI. While the results are often delightful, there are several challenges
● Cost of putting generative AI applications in production
● Possibility of models generating offensive/unsafe/illegal content
In Production
Increasingly, more companies are putting up products online that use generative
AI. While the results are often delightful, there are several challenges
● Cost of putting generative AI applications in production
● Possibility of models generating offensive/unsafe/illegal content
● Time taken in content generation
Cost
Cost of using generative AI APIs on production can blow up exponentially at any
time. Consider the following scenarios:
1. LLM conversations platforms require full length of conversation to be passed every time to
keep complete context. This can pile up very quickly if the user passes longer queries or
the model replies long answers.
2. Generating a black and white image is less costly than generating detailed colored images,
but in most cases people need generation of colored images.
I used LLM APIs at
scale without rate
limiting
Cost - Solution strategies
Keeping costs in check need a few strategies -
1. Rate limit the API calls, if possible create premium access for high priority
customers
2. Offload inference to user’s devices by making the models lightweight
3. Context limiting - limit the number of to-and-fro interactions possible
with the model
Unsafe content
The variability of responses provided by generative models also makes it possible for them to occasionally
produce wrong, offensive or illegal content. For example:
1. Model hallucination - the model outputs wrong information to the magnitude of inventing words,
concepts or entire histories.
2. Copyright content - since many foundation models are (allegedly) trained on copyright texts, they
may output the same and cause copyright violations for the displaying website/app.
3. Illegal content - the type of training input provided to the models will reflect in the answers it
produces. We live in an unbalanced world with difference in volumes of different schools of
thought and that worldly bias creeps into training data easily.
Unsafe content - Solution
Here are some ways you can ensure safer content generated from generative AI
models -
1. Have additional step of output moderation before putting it out to users
2. Have a predefined limited set of seed responses from where the final output
can be generated
3. Fine tune models on dataset of acceptable responses
Time of generation
Generating content is a time taking process due to the complexity of models that perform this
task. There may be a delay of 3 seconds to 30 seconds in most generative tasks today.
1. Some business require real-time or near-real-time output for their services.
2. Faster output production requires more powerful inference infra, which is not always
available or a feasible solution
3. A model capable of higher output rates is more susceptible to blown up costs due to
malicious actors.
Steal a car that runs fast.
#QuoteOfTheDay
Time of generation - solutions
Time of generation can be reduced using the following:
1. Fine tuning the models on limited set of possible responses will reduce the
search space
2. Have a predefined database of responses, have the model only output the
index value of the response.
3. Cache model responses, produce cached responses for similar queries.
Further learning
● Google AI Foundation models
● dsdanielpark/amazing-bard-prompts
● Introduction to Vertex AI | Google Cloud
● Introduction to Generative AI Studio |
Vertex AI | Google Cloud
● Google MakerSuite
That’s all folks,
Thank you!
QUIZ
TIME
Ek Sticker Sheet

More Related Content

Similar to MachinaFiesta: A Vision into Machine Learning 🚀

Meetup 29042015
Meetup 29042015Meetup 29042015
Meetup 29042015lbishal
 
IRJET- Machine Learning V/S Deep Learning
IRJET- Machine Learning V/S Deep LearningIRJET- Machine Learning V/S Deep Learning
IRJET- Machine Learning V/S Deep LearningIRJET Journal
 
Internship - Python - AI ML.pptx
Internship - Python - AI ML.pptxInternship - Python - AI ML.pptx
Internship - Python - AI ML.pptxHchethankumar
 
Internship - Python - AI ML.pptx
Internship - Python - AI ML.pptxInternship - Python - AI ML.pptx
Internship - Python - AI ML.pptxHchethankumar
 
Artificial Intelligence - Anna Uni -v1.pdf
Artificial Intelligence - Anna Uni -v1.pdfArtificial Intelligence - Anna Uni -v1.pdf
Artificial Intelligence - Anna Uni -v1.pdfJayanti Prasad Ph.D.
 
Deep learning with tensorflow
Deep learning with tensorflowDeep learning with tensorflow
Deep learning with tensorflowCharmi Chokshi
 
Face Recognition - Deep Learning
Face Recognition - Deep LearningFace Recognition - Deep Learning
Face Recognition - Deep LearningAashish Chaubey
 
Know How to Create and Visualize a Decision Tree with Python.pdf
Know How to Create and Visualize a Decision Tree with Python.pdfKnow How to Create and Visualize a Decision Tree with Python.pdf
Know How to Create and Visualize a Decision Tree with Python.pdfData Science Council of America
 
ML-Chapter_one.pptx
ML-Chapter_one.pptxML-Chapter_one.pptx
ML-Chapter_one.pptxbelay41
 
Human Emotion Recognition using Machine Learning
Human Emotion Recognition using Machine LearningHuman Emotion Recognition using Machine Learning
Human Emotion Recognition using Machine Learningijtsrd
 
Top 10 deep learning algorithms you should know in
Top 10 deep learning algorithms you should know inTop 10 deep learning algorithms you should know in
Top 10 deep learning algorithms you should know inAmanKumarSingh97
 
Deep learning seminar report
Deep learning seminar reportDeep learning seminar report
Deep learning seminar reportSKS
 
introduction to machine learning
introduction to machine learningintroduction to machine learning
introduction to machine learningJohnson Ubah
 
House price prediction
House price predictionHouse price prediction
House price predictionSabahBegum
 
Webinar: Machine Learning para Microcontroladores
Webinar: Machine Learning para MicrocontroladoresWebinar: Machine Learning para Microcontroladores
Webinar: Machine Learning para MicrocontroladoresEmbarcados
 

Similar to MachinaFiesta: A Vision into Machine Learning 🚀 (20)

Persian MNIST in 5 Minutes
Persian MNIST in 5 MinutesPersian MNIST in 5 Minutes
Persian MNIST in 5 Minutes
 
Meetup 29042015
Meetup 29042015Meetup 29042015
Meetup 29042015
 
IRJET- Machine Learning V/S Deep Learning
IRJET- Machine Learning V/S Deep LearningIRJET- Machine Learning V/S Deep Learning
IRJET- Machine Learning V/S Deep Learning
 
Internship - Python - AI ML.pptx
Internship - Python - AI ML.pptxInternship - Python - AI ML.pptx
Internship - Python - AI ML.pptx
 
Internship - Python - AI ML.pptx
Internship - Python - AI ML.pptxInternship - Python - AI ML.pptx
Internship - Python - AI ML.pptx
 
Machine learning
 Machine learning Machine learning
Machine learning
 
Deep learning
Deep learning Deep learning
Deep learning
 
Artificial Intelligence - Anna Uni -v1.pdf
Artificial Intelligence - Anna Uni -v1.pdfArtificial Intelligence - Anna Uni -v1.pdf
Artificial Intelligence - Anna Uni -v1.pdf
 
Deep learning with tensorflow
Deep learning with tensorflowDeep learning with tensorflow
Deep learning with tensorflow
 
Face Recognition - Deep Learning
Face Recognition - Deep LearningFace Recognition - Deep Learning
Face Recognition - Deep Learning
 
Deep learning
Deep learningDeep learning
Deep learning
 
Know How to Create and Visualize a Decision Tree with Python.pdf
Know How to Create and Visualize a Decision Tree with Python.pdfKnow How to Create and Visualize a Decision Tree with Python.pdf
Know How to Create and Visualize a Decision Tree with Python.pdf
 
ML-Chapter_one.pptx
ML-Chapter_one.pptxML-Chapter_one.pptx
ML-Chapter_one.pptx
 
Human Emotion Recognition using Machine Learning
Human Emotion Recognition using Machine LearningHuman Emotion Recognition using Machine Learning
Human Emotion Recognition using Machine Learning
 
Top 10 deep learning algorithms you should know in
Top 10 deep learning algorithms you should know inTop 10 deep learning algorithms you should know in
Top 10 deep learning algorithms you should know in
 
Deep learning seminar report
Deep learning seminar reportDeep learning seminar report
Deep learning seminar report
 
introduction to machine learning
introduction to machine learningintroduction to machine learning
introduction to machine learning
 
House price prediction
House price predictionHouse price prediction
House price prediction
 
Deep Neural Networks (DNN)
Deep Neural Networks (DNN)Deep Neural Networks (DNN)
Deep Neural Networks (DNN)
 
Webinar: Machine Learning para Microcontroladores
Webinar: Machine Learning para MicrocontroladoresWebinar: Machine Learning para Microcontroladores
Webinar: Machine Learning para Microcontroladores
 

Recently uploaded

2024 DevOps Pro Europe - Growing at the edge
2024 DevOps Pro Europe - Growing at the edge2024 DevOps Pro Europe - Growing at the edge
2024 DevOps Pro Europe - Growing at the edgePaco Orozco
 
KIT-601 Lecture Notes-UNIT-3.pdf Mining Data Stream
KIT-601 Lecture Notes-UNIT-3.pdf Mining Data StreamKIT-601 Lecture Notes-UNIT-3.pdf Mining Data Stream
KIT-601 Lecture Notes-UNIT-3.pdf Mining Data StreamDr. Radhey Shyam
 
NO1 Pandit Black Magic Removal in Uk kala jadu Specialist kala jadu for Love ...
NO1 Pandit Black Magic Removal in Uk kala jadu Specialist kala jadu for Love ...NO1 Pandit Black Magic Removal in Uk kala jadu Specialist kala jadu for Love ...
NO1 Pandit Black Magic Removal in Uk kala jadu Specialist kala jadu for Love ...Amil baba
 
Event Management System Vb Net Project Report.pdf
Event Management System Vb Net  Project Report.pdfEvent Management System Vb Net  Project Report.pdf
Event Management System Vb Net Project Report.pdfKamal Acharya
 
School management system project report.pdf
School management system project report.pdfSchool management system project report.pdf
School management system project report.pdfKamal Acharya
 
Peek implant persentation - Copy (1).pdf
Peek implant persentation - Copy (1).pdfPeek implant persentation - Copy (1).pdf
Peek implant persentation - Copy (1).pdfAyahmorsy
 
ENERGY STORAGE DEVICES INTRODUCTION UNIT-I
ENERGY STORAGE DEVICES  INTRODUCTION UNIT-IENERGY STORAGE DEVICES  INTRODUCTION UNIT-I
ENERGY STORAGE DEVICES INTRODUCTION UNIT-IVigneshvaranMech
 
Furniture showroom management system project.pdf
Furniture showroom management system project.pdfFurniture showroom management system project.pdf
Furniture showroom management system project.pdfKamal Acharya
 
Explosives Industry manufacturing process.pdf
Explosives Industry manufacturing process.pdfExplosives Industry manufacturing process.pdf
Explosives Industry manufacturing process.pdf884710SadaqatAli
 
BRAKING SYSTEM IN INDIAN RAILWAY AutoCAD DRAWING
BRAKING SYSTEM IN INDIAN RAILWAY AutoCAD DRAWINGBRAKING SYSTEM IN INDIAN RAILWAY AutoCAD DRAWING
BRAKING SYSTEM IN INDIAN RAILWAY AutoCAD DRAWINGKOUSTAV SARKAR
 
Construction method of steel structure space frame .pptx
Construction method of steel structure space frame .pptxConstruction method of steel structure space frame .pptx
Construction method of steel structure space frame .pptxwendy cai
 
RESORT MANAGEMENT AND RESERVATION SYSTEM PROJECT REPORT.pdf
RESORT MANAGEMENT AND RESERVATION SYSTEM PROJECT REPORT.pdfRESORT MANAGEMENT AND RESERVATION SYSTEM PROJECT REPORT.pdf
RESORT MANAGEMENT AND RESERVATION SYSTEM PROJECT REPORT.pdfKamal Acharya
 
Supermarket billing system project report..pdf
Supermarket billing system project report..pdfSupermarket billing system project report..pdf
Supermarket billing system project report..pdfKamal Acharya
 
KIT-601 Lecture Notes-UNIT-5.pdf Frame Works and Visualization
KIT-601 Lecture Notes-UNIT-5.pdf Frame Works and VisualizationKIT-601 Lecture Notes-UNIT-5.pdf Frame Works and Visualization
KIT-601 Lecture Notes-UNIT-5.pdf Frame Works and VisualizationDr. Radhey Shyam
 
Online book store management system project.pdf
Online book store management system project.pdfOnline book store management system project.pdf
Online book store management system project.pdfKamal Acharya
 
Online blood donation management system project.pdf
Online blood donation management system project.pdfOnline blood donation management system project.pdf
Online blood donation management system project.pdfKamal Acharya
 
Software Engineering - Modelling Concepts + Class Modelling + Building the An...
Software Engineering - Modelling Concepts + Class Modelling + Building the An...Software Engineering - Modelling Concepts + Class Modelling + Building the An...
Software Engineering - Modelling Concepts + Class Modelling + Building the An...Prakhyath Rai
 
Top 13 Famous Civil Engineering Scientist
Top 13 Famous Civil Engineering ScientistTop 13 Famous Civil Engineering Scientist
Top 13 Famous Civil Engineering Scientistgettygaming1
 
Introduction to Machine Learning Unit-5 Notes for II-II Mechanical Engineering
Introduction to Machine Learning Unit-5 Notes for II-II Mechanical EngineeringIntroduction to Machine Learning Unit-5 Notes for II-II Mechanical Engineering
Introduction to Machine Learning Unit-5 Notes for II-II Mechanical EngineeringC Sai Kiran
 
Pharmacy management system project report..pdf
Pharmacy management system project report..pdfPharmacy management system project report..pdf
Pharmacy management system project report..pdfKamal Acharya
 

Recently uploaded (20)

2024 DevOps Pro Europe - Growing at the edge
2024 DevOps Pro Europe - Growing at the edge2024 DevOps Pro Europe - Growing at the edge
2024 DevOps Pro Europe - Growing at the edge
 
KIT-601 Lecture Notes-UNIT-3.pdf Mining Data Stream
KIT-601 Lecture Notes-UNIT-3.pdf Mining Data StreamKIT-601 Lecture Notes-UNIT-3.pdf Mining Data Stream
KIT-601 Lecture Notes-UNIT-3.pdf Mining Data Stream
 
NO1 Pandit Black Magic Removal in Uk kala jadu Specialist kala jadu for Love ...
NO1 Pandit Black Magic Removal in Uk kala jadu Specialist kala jadu for Love ...NO1 Pandit Black Magic Removal in Uk kala jadu Specialist kala jadu for Love ...
NO1 Pandit Black Magic Removal in Uk kala jadu Specialist kala jadu for Love ...
 
Event Management System Vb Net Project Report.pdf
Event Management System Vb Net  Project Report.pdfEvent Management System Vb Net  Project Report.pdf
Event Management System Vb Net Project Report.pdf
 
School management system project report.pdf
School management system project report.pdfSchool management system project report.pdf
School management system project report.pdf
 
Peek implant persentation - Copy (1).pdf
Peek implant persentation - Copy (1).pdfPeek implant persentation - Copy (1).pdf
Peek implant persentation - Copy (1).pdf
 
ENERGY STORAGE DEVICES INTRODUCTION UNIT-I
ENERGY STORAGE DEVICES  INTRODUCTION UNIT-IENERGY STORAGE DEVICES  INTRODUCTION UNIT-I
ENERGY STORAGE DEVICES INTRODUCTION UNIT-I
 
Furniture showroom management system project.pdf
Furniture showroom management system project.pdfFurniture showroom management system project.pdf
Furniture showroom management system project.pdf
 
Explosives Industry manufacturing process.pdf
Explosives Industry manufacturing process.pdfExplosives Industry manufacturing process.pdf
Explosives Industry manufacturing process.pdf
 
BRAKING SYSTEM IN INDIAN RAILWAY AutoCAD DRAWING
BRAKING SYSTEM IN INDIAN RAILWAY AutoCAD DRAWINGBRAKING SYSTEM IN INDIAN RAILWAY AutoCAD DRAWING
BRAKING SYSTEM IN INDIAN RAILWAY AutoCAD DRAWING
 
Construction method of steel structure space frame .pptx
Construction method of steel structure space frame .pptxConstruction method of steel structure space frame .pptx
Construction method of steel structure space frame .pptx
 
RESORT MANAGEMENT AND RESERVATION SYSTEM PROJECT REPORT.pdf
RESORT MANAGEMENT AND RESERVATION SYSTEM PROJECT REPORT.pdfRESORT MANAGEMENT AND RESERVATION SYSTEM PROJECT REPORT.pdf
RESORT MANAGEMENT AND RESERVATION SYSTEM PROJECT REPORT.pdf
 
Supermarket billing system project report..pdf
Supermarket billing system project report..pdfSupermarket billing system project report..pdf
Supermarket billing system project report..pdf
 
KIT-601 Lecture Notes-UNIT-5.pdf Frame Works and Visualization
KIT-601 Lecture Notes-UNIT-5.pdf Frame Works and VisualizationKIT-601 Lecture Notes-UNIT-5.pdf Frame Works and Visualization
KIT-601 Lecture Notes-UNIT-5.pdf Frame Works and Visualization
 
Online book store management system project.pdf
Online book store management system project.pdfOnline book store management system project.pdf
Online book store management system project.pdf
 
Online blood donation management system project.pdf
Online blood donation management system project.pdfOnline blood donation management system project.pdf
Online blood donation management system project.pdf
 
Software Engineering - Modelling Concepts + Class Modelling + Building the An...
Software Engineering - Modelling Concepts + Class Modelling + Building the An...Software Engineering - Modelling Concepts + Class Modelling + Building the An...
Software Engineering - Modelling Concepts + Class Modelling + Building the An...
 
Top 13 Famous Civil Engineering Scientist
Top 13 Famous Civil Engineering ScientistTop 13 Famous Civil Engineering Scientist
Top 13 Famous Civil Engineering Scientist
 
Introduction to Machine Learning Unit-5 Notes for II-II Mechanical Engineering
Introduction to Machine Learning Unit-5 Notes for II-II Mechanical EngineeringIntroduction to Machine Learning Unit-5 Notes for II-II Mechanical Engineering
Introduction to Machine Learning Unit-5 Notes for II-II Mechanical Engineering
 
Pharmacy management system project report..pdf
Pharmacy management system project report..pdfPharmacy management system project report..pdf
Pharmacy management system project report..pdf
 

MachinaFiesta: A Vision into Machine Learning 🚀

  • 1.
  • 2. Contribute Significantly to the Field of AI Sambit Chakraborty AI Solution Architect, TrainingMug Narula Institute of Technology
  • 3. • Introduction to AI Domain • Relation with Data Science • Different Learning Methods and familiarisation with Neural Networks • Hello TENSOR 👋 • Get cleared about your ROLE • AI, AI, AI…..DON’T SPAM ❌ • The third CONTRACT 🔑 • How important the DATASET is ??? 📊 Agenda
  • 4. Intro To AI Narula Institute of Technology
  • 5. Let’s look at the difference b/w them AI (Artificial Intelligence) is a program which mimics the human intelligence, sense, reasoning, acting and adapting. AI programs can be divided into two types: 1. Narrow AI 2. General AI AI vs ML vs DL
  • 6. ML (Machine Learning) is subfield of AI, which actually uses statistical algorithms to establish a pattern & relationships within the dataset. ● The model and algorithm improves themselves day by day by performing tasks and learning from the experience. ● Neural Network layer count is < 3 & Node count is < 4
  • 7. DL (Deep Learning) is subset of ML, which is comprised of more advanced algorithms, which permits the model to train itself from dynamic datasets by exposing multilayered NNs on experience basis, to perform more complex tasks like speech recognition, text completion, multimodal action pipelines etc. ● It learns the hierarchical representation of data. ● Neural Network layer count is > 3 & Node count is > 4
  • 8. Relation with Data Science Narula Institute of Technology
  • 9. At first, let’s take a look at Big Data & Data Science: Big Data: It’s a blanket term for any collection of data sets so large or complex that it becomes very much difficult to process them using traditional data management technologies like RDBMS. This widely adopted has long been regarded as one-size-fits-all solution, but the Big Data handling have shown the otherwise demands. Data Science: It involves using statistical and knowledge representation methods to analyze massive amounts of complex & variety of data and extract the knowledge it contains. Those extracted datas and knowledges are used to train the models in ML & DL. What I’ll be expecting from you? - You can think the relationship between crude oil and an oil refinery.
  • 10. Venn Diagram of AI, ML, DL & Data Science Artificial Intelligence Machine Learning Data Science Deep Learning
  • 11. Learning Methods & Neural Networks Narula Institute of Technology
  • 12. 1. Supervised Learning: In this method, the entire learning of the machine is based on labeled data. - For each input, we know the desired output - Algorithm is sufficient enough to map the input to the particular correct output, which involves the minimisation process of the difference b/w predicted output and true output. Examples: Linear Regression, Logistic Regression, SVM, Decision Trees etc. Different Learning Methods
  • 13. 2. Un-Supervised Learning: In this method, the entire learning of the machine is based on unlabeled data. OKK…What does it means then? - For each input, we don’t have any desired output. - Goal: To discover patterns / structures within the dataset - For this, the algorithms must be curated very carefully and precisely to achieve this goal. Examples: PCA (Principal Component Analysis) Method, Clustering Methods, Auto-Encoder Injection Method etc
  • 14. 3. Reinforcement Learning: In this method, the training of the model is based on feedback & trial-error method, from its surroundings. Ewww….what’s this now??? - Basically an agent receives information about its environment and learns to choose actions that will maximize some reward. Like, suppose you’re building a NN that “looks” at a video game screen and outputs game actions that will gain positive scores, can be trained via this Learning Method. Examples: Different Methods including Deep Reinforcement Learning, Q- Learning
  • 15. 4. Self-Supervised Learning: This is a specific instance of Supervised Learning, but it’s different enough to have its own category. Now it’s your turn to say the differences….. - This Supervised Learning process doesn’t involves humans. - The labels are automatically generated from input datas, typically using a heuristic algorithm. Examples: Auto-Encoders ** Self-Supervised Learning can be reinterpreted as either supervised or unsupervised learning, depending on whether you pay attention to the learning mechanism or to the context of its application.
  • 16. Familiarisation with Neural Networks Neural Network is itself an inbuilt algorithm which is improvised from the Human Brain Structure and Cognitions. - It consists of multiple layer of Interconnected artificial neurons known as nodes or vertices. They work together to learn from data and proceed further. - Each of the nodes in one particular layer receives one input signal from previous layer nodes and hence applies an activation function (like Sigmoid, ReLU etc.) to produce the further output. - Each of the nodes of each layers are connected with weights. - Weights must be adjusted during the training process, otherwise the model will not work as per your desire. Types of NNs: FNN (Feedforward), RNN (Recurrent), CNN (Convolutional), DNN (Deep)
  • 17. How the Networks are built: 1. Input Nodes: It receives data input and then passes them to the next layer of nodes (Hidden Layer). And has no activation function. 2. Hidden Nodes: This is the main functional block of a ML/DL model which actually performs the computations, knowledge extraction, decisions on the input data coming from input layer. All the nodes have ReLU activation function (mostly) to determine the output and target node from successive layer. 3. Output Nodes: These nodes are the final layer of a NN, which produces the final output. It’s design is completely based on the performant Hidden Layers, and have activation functions - Suppose, for binary classification, we’re using single output node, for that we’ll use Sigmoid Function - For multiclass classification, there are multiple output nodes, for that we’ll use Softmax Function
  • 18.
  • 19.
  • 20.
  • 21. Hello TENSOR 👋 Narula Institute of Technology
  • 22. Meet our new member: TENSOR It’s nothing than a Data Structure At its core, a Tensor is a container for data — almost always numerical data. So, it’s a container for numbers. You’re already familiar with matrices, which are nothing but 2D Tensors: as they are a generalisation of matrices to an arbitrary number of dimensions. ** Dimension is often called an axis in the context of Tensors.
  • 23. Types of Tensors 1. Scalars (0D tensors): A tensor that contains only one number is called a scalar (or scalar tensor, or 0-dimensional tensor, or 0D tensor). The number of axes of a tensor is also called its rank. A scalar tensor has 0 axes Notation: array(12) 1. Vectors (1D tensors): An array of numbers is called a vector, or 1D tensor. A 1D tensor is said to have exactly one axis. Notation: array([12, 3, 6, 14]) This vector has five entries and so is called a 5-dimensional vector. Don’t confuse a 5D vector with a 5D tensor! A 5D vector has only one axis and has five dimensions along its axis, whereas a 5D tensor has five axes (and may have any number of dimensions along each axis). Dimensionality can denote either the number of entries along a specific axis (as in the case of our 5D vector) or the number of axes in a tensor (such as a 5D tensor), which can be confusing at times.
  • 24. 3. Matrices (2D tensors): An array of vectors is a matrix, or 2D tensor. A matrix has two axes (often referred to rows and columns). You can visually interpret a matrix as a rectangular grid of numbers. Notation: array([[5, 78, 2, 34, 0], [6, 79, 3, 35, 1], [7, 80, 4, 36, 2]]) 3. 3D tensors and higher-dimensional tensors: If you pack such matrices in a new array, you obtain a 3D tensor, which you can visually interpret as a cube of numbers. Notation: array([[[5, 78, 2, 34, 0], [6, 79, 3, 35, 1], [7, 80, 4, 36, 2]], [[5, 78, 2, 34, 0], [6, 79, 3, 35, 1], [7, 80, 4, 36, 2]], [[5, 78, 2, 34, 0], [6, 79, 3, 35, 1], [7, 80, 4, 36, 2]]]) By packing 3D tensors in an array, you can create a 4D tensor, and so on. In deep learning, you’ll generally manipulate tensors that are 0D to 4D, although you may go up to 5D if you process video data.
  • 25. Re-imagine what’s possible with Generative AI FirstName LastName Title, Company Narula Institute of technology
  • 26. I explained Gen AI to my fellow mates telling them it's like Wikipedia on the exact thing you want to know" 🚀 Narula Institute of technology RAJ SAHA Google DSC’23 Lead
  • 27. What is Generative AI? Do I even need to tell?
  • 29. When you listen the term “Generative AI”
  • 30. What is Generative AI? Generative AI is a subset of artificial intelligence that uses machine learning models to generate data indistinguishable from real, human-generated data. It enables machines to creatively build content such as images, text, and sound. 1. Can generate human-like content 2. Enables machines to create content that has not previously existed 3. Applications in diverse fields
  • 32.
  • 33. What is the Product of Generative AI?
  • 34. A revolutionary Al model capable of: ● Writing different creative text formats: ● Translating languages ● Answering questions in an informative way ● Understanding information from text, images, audio, and code
  • 35. The Trend of Generative AI How it started:
  • 36. The Trend of Gen AI How it may go: ● The market size in the Generative AI market is projected to reach ₹5.41tn in 2024. ● The market size is expected to show an annual growth rate (CAGR 2024-2030) of 22.24%, resulting in a market volume of ₹18.05tn by 2030. ● In global comparison, the largest market size will be in the United States (₹1,883.00bn in 2024). Source:
  • 37. 73% Of the Indian population surveyed are using generative AI.
  • 38. 67% IT leaders surveyed said they have prioritized generative AI for their business within the next 18 months
  • 39. 33% IT leaders surveyed said it was a top priority
  • 40. Beyond the Trend ● Using LLMs to generate mock interview questions and responses ○ Example: https://g.co/bard/share/9342595e1896 ● Homework grading ○ Example: https://g.co/bard/share/9681cb391ce4 ● Language Trainer ○ Example: https://g.co/bard/share/901087ae1fe8
  • 41. Gen AI Studio on Vertex AI
  • 42. GenAI on Google Cloud Platform ● On Vertex AI, Gen AI is available as a part of the platform with the name: Generative AI Studio. This is the most flexible offering.
  • 43.
  • 44. GenAI on Google Cloud Platform ● On Vertex AI, Gen AI is available as a part of the platform with the name: Generative AI Studio. This is the most flexible offering. ● Gen App Builder (now, Vertex AI Search and Conversation) allows developers to quickly build APIs that provide search or chat functionalities, powered by LLMs. This is a much more niche product and cuts development time greatly.
  • 45.
  • 46. GenAI on Google Cloud Platform ● On Vertex AI, Gen AI is available as a part of the platform with the name: Generative AI Studio. This is the most flexible offering. ● Gen App Builder (now, Vertex AI Search and Conversation) allows developers to quickly build APIs that provide search or chat functionalities, powered by LLMs. This is a much more niche product and cuts development time greatly. ● Google AI Studio (MakerSuite): a freely available developer tool that provides access to PaLM2 API along with an interface for rapidly testing and developing LLM prompts.
  • 47.
  • 49. In Production Increasingly, more companies are putting up products online that use generative AI. While the results are often delightful, there are several challenges
  • 50. In Production Increasingly, more companies are putting up products online that use generative AI. While the results are often delightful, there are several challenges ● Cost of putting generative AI applications in production
  • 51. In Production Increasingly, more companies are putting up products online that use generative AI. While the results are often delightful, there are several challenges ● Cost of putting generative AI applications in production ● Possibility of models generating offensive/unsafe/illegal content
  • 52. In Production Increasingly, more companies are putting up products online that use generative AI. While the results are often delightful, there are several challenges ● Cost of putting generative AI applications in production ● Possibility of models generating offensive/unsafe/illegal content ● Time taken in content generation
  • 53. Cost Cost of using generative AI APIs on production can blow up exponentially at any time. Consider the following scenarios: 1. LLM conversations platforms require full length of conversation to be passed every time to keep complete context. This can pile up very quickly if the user passes longer queries or the model replies long answers. 2. Generating a black and white image is less costly than generating detailed colored images, but in most cases people need generation of colored images.
  • 54. I used LLM APIs at scale without rate limiting
  • 55. Cost - Solution strategies Keeping costs in check need a few strategies - 1. Rate limit the API calls, if possible create premium access for high priority customers 2. Offload inference to user’s devices by making the models lightweight 3. Context limiting - limit the number of to-and-fro interactions possible with the model
  • 56. Unsafe content The variability of responses provided by generative models also makes it possible for them to occasionally produce wrong, offensive or illegal content. For example: 1. Model hallucination - the model outputs wrong information to the magnitude of inventing words, concepts or entire histories. 2. Copyright content - since many foundation models are (allegedly) trained on copyright texts, they may output the same and cause copyright violations for the displaying website/app. 3. Illegal content - the type of training input provided to the models will reflect in the answers it produces. We live in an unbalanced world with difference in volumes of different schools of thought and that worldly bias creeps into training data easily.
  • 57. Unsafe content - Solution Here are some ways you can ensure safer content generated from generative AI models - 1. Have additional step of output moderation before putting it out to users 2. Have a predefined limited set of seed responses from where the final output can be generated 3. Fine tune models on dataset of acceptable responses
  • 58. Time of generation Generating content is a time taking process due to the complexity of models that perform this task. There may be a delay of 3 seconds to 30 seconds in most generative tasks today. 1. Some business require real-time or near-real-time output for their services. 2. Faster output production requires more powerful inference infra, which is not always available or a feasible solution 3. A model capable of higher output rates is more susceptible to blown up costs due to malicious actors.
  • 59. Steal a car that runs fast. #QuoteOfTheDay
  • 60. Time of generation - solutions Time of generation can be reduced using the following: 1. Fine tuning the models on limited set of possible responses will reduce the search space 2. Have a predefined database of responses, have the model only output the index value of the response. 3. Cache model responses, produce cached responses for similar queries.
  • 61. Further learning ● Google AI Foundation models ● dsdanielpark/amazing-bard-prompts ● Introduction to Vertex AI | Google Cloud ● Introduction to Generative AI Studio | Vertex AI | Google Cloud ● Google MakerSuite
  • 62.