> Table of contents

1. Programming

2. Data

3. Statistical models

4. Supervised learning

5. Unsupervised learning

6. Reinforcement learning

7. Neural networks

8. NLP

9. CV

10. GenAI

11. Cloud

12. Hardware

13. Edge

14. Operations

Python programming skills for AI

Techniques, libraries, frameworks, integrations, etc.

1.Python Object Oriented Programming:

OOP makes it easier to organize code and reuse it, just like using one recipe to make many cakes!

• Classes: The blueprint or design for an object (like a recipe for making a cake).

• Objects: The actual thing you create using the class (like the cake you bake from the recipe).

• Attributes: The properties of the object (like the cake's flavor or size).

• Methods: The actions the object can do (like slicing or eating the cake).

Object Oriented Programming crash course:

• Class: A blueprint or template to create objects (e.g., a plan for making cars).

• Object: A real-world thing created from a class (e.g., a specific car built using the plan).

• Attributes: The properties of an object (e.g., a car’s color or brand).

• Methods: The actions an object can do (e.g., a car can drive or honk).

• Encapsulation: Keeping details hidden inside an object, like how a car engine works, so you only focus on using it.

• Inheritance: Creating a new class from an existing one (e.g., a race car class can come from a general car class).

• Polymorphism: Objects can act differently even if they share the same name (e.g., a car honks, and a bike rings a bell, but both are "sounds").

• Abstraction: Hiding unnecessary details and showing only what’s important (e.g., you just drive the car without worrying about the engine parts).

These concepts make code more organized, reusable, and easy to work with!

Key difference between encapsulation and abstraction in OOP:

• Encapsulation: Hides the internal details of an object and protects its data. Example: A car engine is hidden inside the car, and you can’t directly change it.

• Abstraction: Shows only what’s necessary and hides the complexity. Example: You use the car’s steering wheel and pedals without needing to know how the engine works.

Encapsulation is about hiding data, while abstraction is about hiding complexity.

2.Libraries

Here’s a list of popular Python libraries for AI and data science, with brief explanations:

Data Processing:

• NumPy: Handles mathematical operations and arrays efficiently.

• Pandas: Makes it easy to work with structured data (tables).

• Dask: Scales data processing and computation for large datasets.

• Polars: A fast DataFrame library for handling large data efficiently.

• Fireducks: Compiler Accelerated DataFrame Library for Python with fully-compatible pandas API

Visualisation:

• Matplotlib: Creates beautiful charts and plots.

• Seaborn: Simplifies creating advanced visualizations.

• Plotly: Creates interactive visualizations and dashboards.

• Dash: Builds web apps for AI and data visualization using Plotly.

• Bokeh: Creates interactive and browser-friendly visualizations.

• Altair: Declarative library for creating statistical charts.

• Plotnine: A Python library for creating ggplot2-style charts.

• Yellowbrick: Visualizes and evaluates machine learning models.

Statistics:

• Statsmodels: Helps with statistical analysis and modeling.

• Pymc3: Performs probabilistic programming for Bayesian models.

• Scipy: Extends NumPy for scientific computing tasks.

• PyOD: A library for detecting outliers in datasets.

AI model training:

• Scikit-learn: Provides tools for machine learning models and data preprocessing.

• TensorBoard: Visualizes metrics and graphs for TensorFlow models.

• TensorFlow: A powerful library for building and training AI models.

• PyTorch: Another great library for AI, known for flexibility.

• PyTorch Lightning: Simplifies building and training PyTorch models.

• Keras: Simplifies building neural networks using TensorFlow or other backends.

• Keras-RL: A reinforcement learning library built on Keras.

• OpenCV: Used for image and video processing in AI.

• NLTK: Focuses on natural language processing (NLP).

• SpaCy: Another library for NLP, faster and easier to use.

• Gensim: Used for topic modeling and word embeddings in NLP.

• XGBoost: Optimized for building gradient-boosted models.

• LightGBM: Fast and efficient for gradient boosting in large datasets.

• Shap: Explains AI model predictions using visualization.

• FastAPI: Builds APIs to deploy AI models quickly.

• CatBoost: Gradient boosting library for categorical data.

• PyCaret: Simplifies end-to-end machine learning workflows.

• Prophet: Used for time series forecasting.

• DeepFace: A library for face recognition and analysis.

• TensorFlow Lite: A lighter version of TensorFlow for running models on mobile devices.

• Pytorch Geometric: Specialized in deep learning on graph data.

• Gluon: A flexible deep learning library for fast model building.

• LightFM: Used for building recommendation systems.

• Pygame: A library for creating games, often used in AI for reinforcement learning.

• Scrapy: Web scraping framework for gathering large datasets.

• Optuna: Automates hyperparameter optimization for models.

• H2O.ai: Provides machine learning and AI tools for building models.

GenAI:

• Hugging Face Transformers: Specializes in using pre-trained AI models like GPT.

• LangChain

• LangFlow

Operations:

• mlflow: Tracks and manages machine learning experiments.

• Flask:

• Streamlit: Builds interactive web apps for AI projects easily.

• Uvicorn: ASGI (Async. Server Gateway Integration) web server implementation for Python