deep_learning

Deep Learning Cheatsheets

Welcome to the Deep Learning section of the Algorithm Cheatsheets repository!

Deep Learning is a highly specialized subset of Machine Learning that relies on Artificial Neural Networks with multiple layers (hence the term "deep"). Inspired by the structure and function of the human brain, these models are exceptionally powerful at finding highly complex, non-linear patterns in unstructured data like images, audio, and raw text.

Core Concepts

Unlike traditional Machine Learning, which often requires heavy manual "feature engineering" (telling the algorithm exactly what to look for), Deep Learning models perform Representation Learning. They automatically figure out which features are most important as data passes through their hidden layers.

The field is largely divided by architecture types based on the data they process:

1. Standard Tabular Data: Artificial Neural Networks (ANNs)
2. Spatial Data (Images/Video): Convolutional Neural Networks (CNNs)
3. Sequential Data (Text/Time-Series): Recurrent Neural Networks (RNNs) & LSTMs
4. Advanced Sequences (Modern NLP/Vision): Transformers

Directory Contents

Here is what you will find in this directory. Each architecture includes a theoretical Markdown cheatsheet and an accompanying Jupyter Notebook using modern frameworks like TensorFlow/Keras or PyTorch / Hugging Face.

The Foundations

• Artificial Neural Networks (ANN): Cheatsheet
  • Best for: General-purpose deep learning on structured tabular data, or serving as the final output layers for more complex networks.

Computer Vision

• Convolutional Neural Networks (CNN): Cheatsheet
  • Best for: Image classification, object detection, facial recognition, and medical image analysis.

Sequence & Temporal Models

• Recurrent Neural Networks (RNN & LSTM): Cheatsheet
  • Best for: Processing time-series data, speech recognition, and basic sequential text processing where historical memory is required.

The Modern Frontier

• Transformers & Attention: Cheatsheet
  • Best for: State-of-the-art Natural Language Processing (LLMs like GPT/BERT), advanced translation, and increasingly, complex computer vision tasks.

How to use this section

• For Theory: Read the .md files. They contain the mathematical intuition, architectural diagrams, pros and cons, and real-world use cases.
• For Practice: Open the code/ folder to see how to build, compile, train, and evaluate these complex models using industry-standard libraries.
• Note on Hardware: Deep learning models are computationally expensive. We highly recommend running these notebooks on Google Colab, Kaggle, or a local machine with a dedicated NVIDIA GPU.

---

Back to the Main Repository README