AI.info
Unsupervised learning
Finding structure when there is no answer key, and knowing when the structure is in the data or in the method.
- Structure Discovery Without Answer Keys
- Units, Features, and Preprocessing for Unsupervised Data
- Distance, Similarity, and Neighborhoods
- High-Dimensional Geometry and Mixed Data
- Similarity Graphs, Density, and Connectivity
- Evaluating Unsupervised Structure Without Labels
- K-Means: Objective and Lloyd’s Algorithm
- K-Means Initialization, K Selection, and Failure Modes
- Centroid Alternatives and Large-Scale Partitioning
- Hierarchical Clustering and Dendrograms
- Linkage Choices, Connectivity, and Tree Cuts
- DBSCAN: Core, Border, and Noise
- OPTICS and HDBSCAN Across Density Scales
- Gaussian Mixtures, EM, and Soft Membership
- Spectral Clustering and Graph Cuts
- BIRCH, Streaming Summaries, and Massive Data
- Categorical and Mixed-Type Clustering
- Constrained and Semi-Supervised Clustering
- Biclustering and Co-Clustering
- Text and Embedding Clustering
- Time-Series Clustering and Alignment
- Spatial and Geospatial Clustering
- Graph Community Detection and Network Partitions
- Interpreting, Naming, and Profiling Clusters
- Stability, Consensus, and Reproducibility
- Production Assignment, Drift, and the Cluster Lifecycle
- Anomaly Detection: Scope, Reference, and Rarity
- Local Outlier Factor and Neighborhood Anomalies
- Isolation Forest and One-Class Boundaries
- Time-Series Anomalies and Change Points
- Anomaly Thresholds, Alert Budgets, and Evaluation
- PCA: Variance, Components, and SVD Intuition
- PCA in Practice: Scaling, Whitening, and Reconstruction
- NMF, ICA, and Interpretable Factorizations
- Random Projections, Feature Agglomeration, and Compression
- Manifold Learning: MDS, Isomap, LLE, and Diffusion Maps
- t-SNE: Local Neighborhood Visualization
- UMAP: Graph-Based Embeddings and Tradeoffs
- Projection Traps, Trustworthiness, and Joint Workflows
- Capstone: Build and Defend an Unsupervised System