AI Essentials

CIS 9793
Deep Learning
 Agenda
Today’s Session – Week 3: Basics of Machine Learning

1. AI Current Events
2. The Basics of Machine Learning
1. ML Operations
2. Supervised
3. Unsupervised
4. Reinforcement
5. Human Equivalent

3
AI Current Events
The Basics of Machine Learning
 Machine Learning Operations

Business Analysis Data Analytics

Understand the
Evaluate the
business problem & Prepare the Data Choose a Model Train the Model
Collect Data Model
• Data Quality!! • Data Cleansing • Supervised • Feed Training Set • Feed Testing Set
• Data Wrangling • Unsupervised • Do not feed it the
• Data Transformation • Reinforcement training set!!

• Data Visualization *Semi-Supervised

• Develop Training Set **We will circle back to
& Testing Set this – choosing the best • Based on the accuracy
• Data Quality!!! model depends on the of your model’s
business needs. In for predictions, decide if
example, if we want AI to you can adjust any of
provide us predictions vs the variables of your
categorizations certain model to increase its
models (i.e. algorithms) accuracy.
are better suited than
others. Deploy The Model Parameter Tuning

6
 Machine Learning vs Coding
Machine Learning vs. Traditional Coding (aka Programming, Modeling, etc.)

• Traditional Programming involves writing explicit instructions for a computer to execute specific tasks.
• Machine Learning, on the other hand, allows systems to learn and improve from data without explicit programming.
• In ML, algorithms are trained on data to make predictions or decisions.

https://sravya-tech-usage.medium.com/traditional-programming-vs-machine-
learning-e9bbed5e491c

https://www.slideteam.net/machine-learning-vs-traditional-programming-ppt-powerpoint-presentation-
icons.html#images-1

7
 Supervised Learning

List of Common Algorithms

• Linear Regression
• Logistic Regression
• Decision Tree
• Support Vector Machines
• Neural Networks

What is it?

• Human experts acts as the teacher where we feed the

computer with training data containing the input/predictors
and we show it the correct answers (output) and from the
data the computer should be able to learn the patterns.

• Supervised learning algorithms try to model relationships

https://www.geeksforgeeks.org/supervised-unsupervised-learning/ and dependencies between the target prediction output and
the input features such that we can predict the output values
for new data based on those relationships which it learned
from the previous data sets.

8
 Unsupervised Learning

List of Common Algorithms

• K-Means Clustering
• Association Rules

What is it?

• We feed data to the machine and let it teach itself through

pattern recognition. The computer might be able to teach
you new things after it learns patterns in data.

• It is mainly used in pattern detection and descriptive

modeling. However, there are no output categories or labels
here based on which the algorithm can try to model
relationships. These algorithms try to use techniques on the
https://learnandbuildspace.quora.com/What-is-unsupervised-learning
input data to mine for rules, detect patterns, and summarize
and group the data points which help in deriving meaningful
insights and describe the data better to the users.

9
 Reinforcement Learning
Reinforcement Learning List of Common Algorithms

• Q-Learning
• Temporal Difference
• Deep Adversarial Networks

What is it?

• Uses observations gathered from the interaction with the

environment to take actions that would maximize the reward
or minimize the risk. Reinforcement learning algorithm
(called the agent) continuously learns from the environment
in an iterative fashion. In the process, the agent learns from
its experiences of the environment until it explores the full
range of possible states.

• Reinforcement Learning is a type of Machine Learning, and

thereby also a branch of Artificial Intelligence. It allows
https://www.geeksforgeeks.org/understanding-reinforcement-learning-in-depth/
machines and software agents to automatically determine the
ideal behavior within a specific context to maximize its
performance. Simple reward feedback is required for the
agent to learn its behavior; this is known as the
reinforcement signal.
10
 Human Equivalents
Common Machine Learning Human Equivalent

Supervised Learning:

Human Equivalent: It's like teaching a student by providing answers and questions in advance. The student learns to predict the right answers based
on the questions and answers provided.

Unsupervised Learning:

Human Equivalent: Imagine sorting a pile of various fruits without knowing their names. You start noticing patterns and group similar fruits
together, even though you don't know what each fruit is called.

Reinforcement Learning:

Human Equivalent: Think of a dog being trained with treats. When it performs a desired action (good action), it receives a treat (reward). If it
misbehaves (bad action), it gets a scolding or no treat (penalty). Over time, the dog learns to perform more good actions.

11
 Common ML Algorithms
Common Machine Learning Algorithms

• Linear Regression: Predicts a continuous outcome based on input features.

• Decision Trees: Hierarchical tree-like models used for classification and regression.

• Random Forest: An ensemble learning method based on decision trees.

• Support Vector Machines (SVM): Classify data by finding a hyperplane that best separates classes.

• K-Means Clustering: Unsupervised algorithm for clustering data points into groups.

• Neural Networks: Composed of interconnected artificial neurons (perceptrons).

• Convolutional Neural Networks (CNNs): Specialized for image processing tasks, using convolutional layers to detect features.

• Recurrent Neural Networks (RNNs): Suitable for sequential data, using recurrent connections to maintain information about previous inputs.

• Image Recognition: CNNs are used in image classification, object detection, and facial recognition.

• Natural Language Processing (NLP): RNNs and LSTM networks power chatbots, language translation, and sentiment analysis.

• Autonomous Vehicles: Deep learning is crucial for self-driving cars to perceive their environment.

• Recommendation Systems: RNNs and CNNs can personalize recommendations for users.

12
 Human Equivalents
Common Machine Learning Algorithm Human Equivalent

Linear Regression:

Human Equivalent: Predicting a student's final grade based on their study hours. You notice a pattern: as study hours increase, grades tend to
improve.

Decision Trees:

Human Equivalent: Creating a flowchart to help you decide what to do on a rainy day. You start with a question like "Is it raining?" and then follow
branches to decide whether to stay in or go out.

Random Forest:

Human Equivalent: When making a tough decision, you consult multiple friends with different expertise. You gather their opinions and make your
decision based on their collective advice.

Support Vector Machines (SVM):

Human Equivalent: Finding the best way to separate a group of people into two teams based on their interests and skills. You look for the most
effective line that keeps the teams as different as possible.

K-Means Clustering:

Human Equivalent: When organizing a library, you group similar books together without knowing their titles or content. You use common
characteristics like size, cover color, or topic.

13
Q&A
End
CIS 9793