AI/DEEP LEARNING

WITH PYTHON COURSE

By :Ms. Shyma Mogtaba

All right reserved for Intelligent Pattern co.Ltd

 WELCOME GUYS
TOPRESENTATION
IamShymaMogtaba
DataScientist
You can find me at :
Email: [email protected]
LinkedIn:
https://www.linkedin.com/in/shyma-
uthman-691133152

LETS GETSTARTED
 LET'S KNOW EACH
OTHER...

Mayeveryone introduce himself?

NEXT...
03 INTRODUCTION TO AI
AGENDA

04 AI BENEFITS & CHALLENGES

The presentation will gothrough
the following...

01
ABOUT US

05
HOW AI WORKS

06 AI LEARNING APPROACHES 02 INTRODUCTION TO DATASCIENCE

09 NN MODELS &ARCHITECTURES
AGENDA

10 BUILD AI SYSTEM
The presentation continue to
cover:

07
AI FUNCTIONAL DESIGNS

11 AI APPLICATIONS & BEST

PRACTICES.

12 AI DESIGN PATTERNS.
08 INTRODUCTIONTONEURAL
NETWORKS
 EXPLAINING
ALL ABOUTUS
WHO WE ARE?
Intelligent Pattern aimed to be amongst leading digital transformation
solutions provider, providing Data Science training, Intelligent
Solutions and Consultancy Services to companies and organizations in
Sudan.
Intelligent Pattern a business name with aim of bringing data to life in
Sudan, providing training & workshops programs for public,
researchers, students, fresh graduates, organizations and everyone
interested in Data Science in Sudan, developing & optimizing Data
Science models & solutions to real world problems, providing
consultancy services to organizations, and creating our own products
value (Artificial Intelligence, Machine Learning ,CRM,
(Natural Language Processing (NLP), Big Data Analytics, Robotic Process
Automation (RPA), BI and many Data Science Solutions ).
Our experts team brings a unique set of skills and experience in the field
of Data Science.
INTRODUCTION TO
DATA SCIENCE
“Data Science”: it is fundamentally an
interdisciplinary subject. Data science comprises
three distinct and overlapping areas: the skills of a
statistician who knows how to model and summarize
datasets (which are growing ever larger); the skills of
a computer scientist who can design and use
algorithms to efficiently store, process, and visualize
this data; and the domain expertise—what we might
think of as “classical” training in a subject—necessary
both to formulate the right questions and to put their
answers in context.
ARTIFICIAL
I NTELLI GENCE
Artificial Intelligence is becoming
increasingly relevant in the modern world
where everything is driven by technology
and data. It is used extensively across many
fields such as search engines, image
recognition, robotics, finance, and others.
Deep Learning techniques working with
actual data, developing algorithms using
Neural Network, performing operations and
constructing model.
 AI, ML AND DL

Whatis the relationship

betweenAI,Ml and
Dl???
UNDERSTANDING
AI Existing knowledge and reference
to compare against

Memory Learning

New knowledge to memorize

BENEFITS OFAI

Informed, Accurate and Responsive

Decision Making
Responsive Business Solutions
Self-Learning and Adaptive Solutions
Autonomous Solutions
Automated Management
Emulating Human Behavior
Identifying and Responding to Malicious
Behavior
Solving Problems of Higher Complexity
COMMON AI Learning Curve and Incomprehensible Decision
CHALLENGES Making
Distrust by Humans
MOST COMMON AI CHALLENGES Data Hosting and Access Restrictions
Many challenges faces AI such as : Quality Assurance
 HOW AI WORKS

Training results
Training data Model in training

The analyst provide old data used to train systems models and then evaluates the results
many iterations may occure before the model is considered sufficiently trained
 HOW AI WORKS

Production data

AI system

Model in training Trained model

Training results

Once the model has been trained its deployed for use with few input data relevant
to the problem.The results produced can be provided to the analyst or can be Analyst
fed back into AI system
 HOW AI WORKS
Make decisions the
Self learning first time.

Learn from outcome and

make improved decisions.

Production data Trained model Training results AI system

Learn from outcome and
make even more improved
decisions.
The AI system can repeatedly lookup and analyze previously
inputdata inorder to learn from prior outcomes so that it can etc...
improve future outcomes.
 AI LEARNING
APPROACHES
SUPERVISED LEARNING
Classification, Regression...etc

UNSUPERVISEDLEARNING

Clustering, segmentation ...etc

CONTINUOUS LEARNING
Used in real-world SEMI SUPERVISEDLEARNING
environment
Mixed approach

HEURISTIC LEARNING
Short time solutions, speed up REINFORCMENT OF MACHINE
decision making, flexibility LEARNING
Step by step
AI FUNCTIONAL
DESIGN
Computer Vision
Pattern Recognition
Robotics
NLP & Speech Recognition
Frictionless Integration
Fault Tolerance Model Integration
INTRODUCTION TO
NEURAL NETWORKS
An artificial neural network is a system designed to
emulate the structure and function of the human
brain in order to give machines the ability to learn
and to make intelligent decisions and, in some cases,
to learn from decisions.
 NN & MODELS

An artificial neural network exists as a compiled software program with one or more
repositories that store data asparameters relevant to its processinglogic.
It has neurons (referred to as nodes or cells) that exist as mathematical functions within its
softwarelogic.
In both machine learning and AI, a primary objective is to train a model so that, subsequent to
being subject to training data, the system is optimized to process production data for real-
world decision-making purposes.
It is important to understand that for an AI system that uses a neural network, the neural
network is the model. Training the model is synonymousto training the neuralnetwork.
NN COMPONENTS
A neural network is primarily comprised of three components:
neurons
layers
Functions
The neurons are organized into the following three primary layers:
input
hidden
output
The neurons are connected to form links to form the network of
neurons. Data is passed between neurons across links within a
layer or across layers.
ACTIVATION FUNCTION
Activation functions (also known as transfer
functions), provide basic data processing routines
Input Hidden Output
that can be assigned to neural cells.
Common activation functions include: Sigmoid X X
Sigmoid
Tanh Tanh X X
ReLU
ReLU X
Leaky ReLU
Softmax Leaky ReLU X
Softpluss
Softmax X

Softplus X
NEURONS CELLTYPES
A Input Cell
Backfed Input Cell
Noisy Input Cell
Hidden Cell
Probabilistic Hidden Cell
Spiking Hidden Cell
Recurrent Cell
Memory Cell
Recurrent Memory Cell
Kernel Cell
Convolution Kernel or Pool Cell
Output Cell
Match Input/Output Cell
NEURAL NETWORK
ARCHITECTURES
Fundamental Neural Network Architectures are :
Perceptron (P)
Feedforward (FF)
Radial Basis Forward (RBF)
Deep Feedforward (DFF)

-(FF) Solve many regression &

classification problems.
- (DFF) Speech recognition.
NEURAL NETWORK
ARCHITECTURES
Architectures with Encoding and Decoding Layers
AutoEncoder (AE)
Sparse AutoEncoder (SAE)
Variational AutoEncoder (VAE)
Denoising AutoEncoder (DAE)

-Reduce dimensions before used in

classification.
NEURAL NETWORK
ARCHITECTURES
Recurrent Cell-based Architectures are:
Recurrent Neural Networks (RNNs)
Echo State Network (ESN)

-(RNN Remember data overtime.

-Past decisions influence current decision.
-Learning sequential data.
NEURAL NETWORK
ARCHITECTURES
Memory-Influenced Architectures
Long/Short Term Memory (LSTM)
Neural Turing Machine (NTM)

- (LSTM) Use memory to learn

overtime.
- Apply to learn from historical data.
- Play chest, video game.
NEURAL NETWORK
ARCHITECTURES
Parabolicity Cell-Driven Architectures
Boltzmann Machine (BM)
Restricted Boltzmann Machine (RBM)
Deep Belief Network (DBN)
Markov Chain (MC)
- (BM) Used when no output required,
monitor sensor data to report
abnormality.
- (RBM) Used to determine preferences
& recommendations
- (DBN) Used for image recognition,
faster processing
NEURAL NETWORK
ARCHITECTURES
Pool and Kernel Influenced Architectures
Deep Convolutional Network (DCN)
De-convolutional Network (DN)
-(DCN) Reduce dimensions,(automate identification of
features should be removed), image analysis
NEURAL NETWORK ARCHITECTURES
Hidden Cell Connection Architectures
Extreme Learning Machine (ELM)
Deep Residual Network (DRN)
Support Machine Vector (SVM)
Kohonen Network (KN) –(Self Organizing Map)
-(ELM) Used when less data available, real time learning tasks, high speed
reconnection
-(DRN) Flexibility to add/remove layers, learn faster
-(SVM) handle large datasets, binary/multi classification, control overfitting
-(KN) Used for unsupervised classification, image recognition, reduce dimensionality
NEURAL NETWORK
ARCHITECTURES
Other Architectures
Hopfield Network (HN)
Generative Adversarial Network (GAN)
Liquid State Machine (LSM)
- (HN) Used for small/incomplete dataset
- (GAN) Image reproduction, generate real life images
- (LSM) Used for computer vision, speech recognition
WHAT AI USEDFOR
Complex Analysis
Graphical Recognition
Language and Sentiment Analysis
Predictions and Forecasting
Autonomous Decision Making
etc..
BULDING AI
SYSTEM
THE BESTPRACTICE
Educate & Make AccurateDecision
Create Use Cases
Focus to Human Element
Prepare for Limitations
Continue to Refine
AI DESIGNING
PATTERN
Data Wrangling Patterns
Feature Encoding
Feature Imputation
Feature Scaling
Text Representation
Dimensionality Reduction
AI DESIGNING
PATTERN
Supervised Learning
Supervised Network Configuration
Image Recognition
Sequence Identification
Unsupervised Learning
Pattern Identification
Content Filtering
AI DESIGNING
PATTERN
Model Evaluation Patterns
Training Performance Evaluation
Prediction Performance Evaluation
Baseline Evaluation
Model Optimization Patterns
Overfitting Avoidance
Frequent Model Retraining
Transfer Learning
PACKAGES OFFEREDBY
INTELLIGENTPATTERN
TRAINING
We provid training in data science, data
analysis and business intelligence

SOLUTIONS
We provide intelligent solutions including intelligent
models customized with your data.

CONSULTANCY
We provide consulting services
for business growth.
 OURSERVICES

TRAINING SOLUTIONS CONSULTANCY

Python for Data Science Crash Course Data Analytics Model Design solution proposal.
Data Analytics course using Python BI Model. Current situation of business analysis and
and R. CRM Model. problem formation.
Business Intelligence. AI / ML Model. Solution Co-implementation.
Data Mining. Text Analytics & NLP Model. Enabling Data Science revolution
Machine Learning. Sentiment Analysis Model. for business growth.
Artificial Intelligence/Deep Learning .
Python, Power BI, RapidMiner, Excel
Robotic Process Automation (RPA).
and More...
MACHINE LEARNING
WITH PYTHONCOURSE
MODULE 1:
01
AI Fundamentals.

MODULE 2: MODULE5:
02 Introduction to Deep Learning. 05
RNN Models

MODULE6:
MODULE 3:
03 06 CNN Models
Nurel Netwotks.

MODULE4:
04 NN Models
MEET OURTEAM

SHYMAMOGTABA
FOUNDER, DATASCIENTIST
Seasonal Data Scientist with 5+ years’ hands-on expertise in Data Mining, application of Artificial Intelligence
and Machine Learning techniques to solve real world problems. University lecturer delivering courses in
Database, Artificial Intelligence, Machine Learning, Expert system, and Data Mining. Holds BSc. Computer
Science, MSc. Computer Science and certified Artificial Intelligence Specialist and Trainer.
LinkedIn: https://www.linkedin.com/in/shyma- uthman-691133152
FIND & CONTACT
US ON DETAIL
[email protected]
+249926186202
+966536261489

Impact hubKhartoum, Riyad117

street

Intelligent Pattern

Intelligent Pattern
 REFERENCES:
https://www.linkedin.com/learning/paths/become-a-data-analyst
Python Data Science Handbook by Jake Vander Plas Copyright ©2017 Published by
O’Reilly Media
Python for Data Science and Machine Learning Bootcamp
https://www.udemy.com/the-data- science-course-complete-data-science-bootcamp
Python Data Science Handbook by Jake Vander Plas Copyright ©2017 Published by
O’Reilly Media
Introduction to Statistical Learning with Applications in R Second Edition by Gareth
James , Daniela Witten , Trevor Hastie , Robert Tibshirani 2021
Arcitura Institution AI/ML course materials
 THANK YOU
FOR YOUR TIME