> ML_LITERATURE // FISHER-1936-USE-OF-MULTIPLE-MEASUREMENTS_v1.0
The Use of Multiple Measurements in Taxonomic Problems
Ronald A. Fisher · Annals of Eugenics (1936)
foundational1936foundationalthirdPartyReproduced
Principal Contribution
Introduced Linear Discriminant Analysis (LDA) and the iconic Iris dataset for statistical classification.
Operational Relevance
Directly applicable for linear supervised dimension reduction and fast, low-compute baseline classification in embedded systems.
Assumptions
- Classes share a common covariance matrix
- Predictor features follow multivariate normal distributions within classes
Limitations
- Assumes identical covariance across all classes, failing when covariance structures diverge (requires QDA)