Cheatsheet

ARTIFICIAL
INTELLIGENCE
Connect with Alumni
 Artificial Intelligence

Artificial Intelligence (AI) refers to the development of computer systems or

algorithms that can perform tasks that typically require human intelligence.
These tasks include learning, reasoning, problem-solving, perception,
language understanding, and decision-making.

Deep Learning Expert Systems NLP

Robotics Computer Vision

Artificial Intelligence
Key Branches
Neural Networks Reinforcement Learning

0101
Speech Recognition 0101 Data Mining Machine Learning

02 AI Cheatsheet
 Machine Learning Deep Learning Machine Learning

Subset of Al that focuses on Specialized form of ML involving Branch of Al that empowers

developing algorithms enabling neural networks with multiple machines to understand,
systems to learn from data and layers, capable of learning interpret, and generate human
improve their performance complex patterns and language, facilitating
over time without explicit representations, often used for communication and interaction
programming. tasks like image and speech between computers and users.
recognition.

Computer Vision Reinforcement Learning

Involves teaching machines to Al paradigm where agents learn

interpret and make decisions by interacting with an
based on visual data, enabling environment, receiving feedback
tasks such as image recognition, in the form of rewards or
object detection, and scene penalties, and adjusting their
understanding. behavior to maximize cumulative
rewards over time.

Did you know?

Machine Learning is the parent, and Deep Learning is its specialized

offspring, showcasing the intricate layers of intelligence in the world
of artificial intelligence.

03 AI Cheatsheet
 Al Algorithms

ML Algorithms DL Algorithms

Linear regression, SVM, Decision

Regression
tree regression, Lasso regression
Convolutional Neural Networks
Supervised (CNNs)
Specialized for image recognition,
Logistic regression, Naive Bayes, using convolutional layers to detect
Classification Decision tree regression, features hierarchically.
K-nearest neighbors

Recurrent Neural Networks (RNNs)

Regression
Designed for sequence data,
capable of capturing temporal
Supervised dependencies in tasks like natural
language processing and speech
Classification recognition

Long Short-Term Memory (LSTM)

A type of RNN with memory cells,

facilitating better handling of
long-range cependencies in
secuential data.

04 AI Cheatsheet
 2 Artificial Intelligence Frameworks

NLTK

scikit

Quality data is the lifeblood of artificial intelligence, as the accuracy and

effectiveness of Al models heavily depend on the richness, relevance, and
cleanliness of the data they are trained on. Robust datasets not only enhance
model performance but also mitigate biases, ensuring more reliable and
ethical Al outcomes.

05 AI Cheatsheet
 3 Data preprocessing techniques

1 2 3

Rescaling Data Mean Removal Standardizing Data

6 5 4

Label Encoding Normalizing Data One Hot Encoding

Binarizing Data

06 AI Cheatsheet
 4 Steps in training an Al model

Data Gathering Data Gathering

& & Training
Preprocessing Preprocessing

Final Model Evaluation &

Development Hyperparameter
& Deployment Tuning

Steps involved in an EDA Evaluation Metrics

Data collection

Finding all Variables and

Understanding Them Precision &
Accuracy F1 - Score
Recall

Cleaning the Dataset

Evaluation Metrics

Identify Correlated Variables

Mean-Squared Mean-Absolute
R-squared
Error Error
Choosing the Right
Statistical Methods

Visualizing and Analyzing

Results

07 AI Cheatsheet
 Underfitting vs Overfitting

Under-fitting Appropriate-fitting Over-fitting

"Don't overfit the data, which means tailoring your model too closely to the
training set. It might work well on what you've seen before but may struggle
with new data. On the flip side, don't underfit, as that's like having a model
that oversimplifies the patterns, missing out on the details in your data."

5 Natural Language Processing

Tokenization Stop Words Stemming

Splitting text into individual Common words (e.g., "and", Reducing words to their base or
words or tokens. "the") often removed during root form (e.g., "running" to
preprocessing "run"").

08 AI Cheatsheet
 Lemmatization POS Tagging Named Entity Recognition (NER)

Similar to stemming but ensures Assigning parts of speech (noun, Identifying named entities
resulting words are valid verb, adjective, etc.) to each (e.g., person names, locations)
(e.g., "went" to "go"). word in a sentence. in text.

Word Embeddings

Representing words as dense

vectors in a continuous
vector space.

NLP Tasks

Language generation Answering questions Text classification

Sentiment analysis Machine translation

09 AI Cheatsheet
 Top 10 Pre-Trained LP Language Models for Al Application Building

1 BERT 2 ROBERTa 3 OpenAl's GPT-4

4 ALBERT 5 XLNet 6 OpenAl's GPT-3

7 StructBERT 8 T5 9 ELECTRA

10 DeBERTa

6 Computer Vision

10 AI Cheatsheet
 Image Representation Image Processing Feature Extraction

Images are Technques to manipulate Identifying relevant

represented as arrays of and enhance images features in images (e.g.,
pixels values (grayscale (e.g., filtering, edges, corners and
or RGB) transformations). textures).

Object Detection Sementic Segmentation Image Classification

Locating and Classifying Assigning class labels to Assigning a label or

objects within images each pixel in an image. category to an entire
image.

Reinforcement Learning in ML

Input Raw Data Environment Output

Reward Best action

Selection of
State
algorithm

Agent

11 AI Cheatsheet
 AI models in real-world applications

Text Generator Image Generator Image Generator

Promt Promt

Generate
Generate
A

Sentiment Analysis Entity Recognition Quiz

Person

Location Product Correct Answer

Considerations for Model Deployment

Successful model deployment demands through testing, scalability planning and

ethical considerations to ensure effective integration and sustained performance
in real-world applications

12 AI Cheatsheet
 Ethical considerations in Al encompass addressing biases embedded in
algorithms to prevent discriminatory outcomes and prioritizing fairness
and transparency to ensure accountable and equitable decision-making
in the development and deployment of Al systems. Upholding these
principles is essential for building trustworthy and socially responsible
artificial intelligence.

Online Platform for AI Development

AI Governance and Regulation

AI governance and regulation involve establishing framework to oversee the ethical

development, deployment , and impact of artificial intelligence, ensuring responsible
use and safeguarding against potential risks to privacy, security and societal
well-being. Balancing innovation with accountability, these measures aim to guide
the ethical evolution of AI technologies.

13 AI Cheatsheet
Data Science Course - Mastering the Python and SQL for Data Science
Fundamentals

Customer Churn Prediction Boston House Price

Project in Machine Learning Prediction

Register Now