Concepts Library

Browse all machine learning concepts organized by category. Each concept page provides focused, in-depth explanation of a single idea.

Neural Network Foundations

Core concepts for understanding neural networks:

• Linear Classifiers - SVM, softmax, and linear decision boundaries
• Perceptron - Single neuron architecture
• Multi-Layer Perceptrons - Deep feedforward networks
• Backpropagation - Gradient computation via chain rule
• Optimization - SGD, momentum, Adam, learning rates
• Regularization - L2 weight decay, early stopping, data augmentation
• Dropout - Stochastic regularization technique
• Bias-Variance Tradeoff - Model capacity and generalization
• Training Practices - Weight initialization, debugging, hyperparameters

Computer Vision

Concepts for visual understanding:

• Convolution - Convolution operation for CNNs
• Pooling - Spatial downsampling (max, average, global)
• Transfer Learning - Pre-training and fine-tuning strategies

Attention and Transformers

Modern sequence modeling with attention:

• RNN Limitations - Why RNNs struggle with long sequences
• Attention Mechanism - Core attention formulation
• Scaled Dot-Product Attention - The transformer attention formula
• Multi-Head Attention - Parallel attention heads
• Transformer Training - Masking strategies, label smoothing, warmup

Language Models

Autoregressive language modeling with GPT:

• GPT Architecture - Decoder-only transformer for language
• Causal Attention - Masked attention for autoregressive generation
• Tokenization - BPE and subword tokenization
• Language Model Training - Training techniques for LMs
• Text Generation - Greedy, top-k, top-p, beam search
• Language Model Scaling - Scaling laws and compute-optimal training

Multimodal Learning

Vision-language models and multimodal fusion:

• Multimodal Foundations - Fusion strategies and alignment
• Contrastive Learning - InfoNCE loss and self-supervised learning
• Advanced Vision-Language Models - Flamingo, BLIP-2, LLaVA

Generative Models

Image and content generation:

• Generative Models - GANs, VAEs, and diffusion comparison
• Diffusion Fundamentals - Forward process, noise schedules, sampling
• Classifier-Free Guidance - Text-to-image conditioning

Advanced Training Topics

Modern training techniques and dynamics:

• Self-Supervised Learning - Pre-training without labels
• Masked Prediction - BERT, MAE, and masked learning
• Training Dynamics - Double descent and overparameterization
• Practical Training Techniques - Warmup, gradient clipping, mixed precision

Browse by Difficulty

Beginner

Foundational concepts for getting started:

• Linear Classifiers
• Perceptron
• Backpropagation
• Convolution
• Pooling

Intermediate

Core deep learning techniques:

• Multi-Layer Perceptrons
• Optimization
• Attention Mechanism
• GPT Architecture
• Multimodal Foundations

Advanced

Cutting-edge techniques and theory:

• Transformer Training
• Language Model Scaling
• Training Dynamics
• Advanced Vision-Language Models
• Practical Training Techniques

Explore More

• Papers Library → - Key research papers analyzed
• Examples → - Hands-on implementations
• Blog → - Applications and industry insights
• Learning Paths → - Structured learning journeys