The taxonomy below reflects the current AI landscape, where models often overlap categories (e.g., CNNs are used in both computer vision and generative art). Model selection depends on factors like data type, task complexity, and desired output format, with newer architectures frequently blending multiple approaches for enhanced performance.
I’m documenting this page as part of my learning process, categorizing AI models based on learning methodologies and architectures.
Learning Methodologies
- Supervised Learning
- Unsupervised Learning Definition: Models discover patterns in unlabeled data by clustering or reducing dimensionality (e.g., customer segmentation).
- Reinforcement Learning
Definition: Models learn using labeled datasets to map input to known outputs (e.g., flower classification).
Definition: Models learn by interacting with an environment and maximizing rewards.
Deep Learning Architectures
Multi-layered neural networks designed for complex pattern recognition and learning from large-scale data:
- Convolutional Neural Networks (CNNs): Specialized for image and video processing
- Recurrent Neural Networks (RNNs): Ideal for sequential data analysis (e.g., time series or text)
- Long Short-Term Memory (LSTM): Designed to handle long-range dependencies in sequential data
- Deep Belief Networks (DBNs): Layered probabilistic models for hierarchical data representation
Specialized Model Types
- Naive Bayes: Probability-based model for tasks like text classification and spam detection
- Learning Vector Quantization (LVQ): Prototype-based classification technique for separating data classes
- Hybrid Models: Combine both predictive and generative capabilities, offering flexibility for tasks like recommendation systems or AI creativity
Get in touch if you have more to add and /or share it if you feel it’s useful!