Examples

Home | Installation | Implementation | Examples | Benchmarks | API | NN Methods | Backends | References

KNN On Iris

library(faissR)

x <- scale(as.matrix(iris[, 1:4]))
labels <- iris$Species

nn_res <- nn(x, k = 15, backend = "auto", metric = "euclidean", n_threads = 4)
head(nn_res$indices)
head(nn_res$distances)

Non-Euclidean Metrics

Cosine and correlation use validated exact CPU paths, FAISS Flat/IVF/IVFPQ CPU/GPU routes through normalized inner-product search, and FAISS CPU HNSW when selected. Inner-product search is available for exact CPU scoring and validated FAISS IP-capable routes where supported.

knn_cos <- nn(x, k = 15, backend = "auto", metric = "cosine", n_threads = 4)
knn_ip <- nn(x, k = 15, backend = "cpu", method = "flat",
             metric = "inner_product", n_threads = 4)

kNN Classifier

set.seed(1)
train <- sample(seq_len(nrow(x)), 100)
test <- setdiff(seq_len(nrow(x)), train)

fit <- knn(x[train, ], labels[train], backend = "auto", metric = "euclidean")
pred <- predict(fit, x[test, ], k = 5)
mean(pred == labels[test])

prob <- predict(fit, x[test, ], k = 5, type = "prob")
head(prob)

# Fit and predict in one call
pred2 <- knn(x[train, ], labels[train], x[test, ], backend = "auto", k = 5)

k-means

fast_kmeans() exposes CPU/FAISS/CUDA/cuVS k-means-style clustering routes [7-8].

km <- fast_kmeans(x, centers = 3, backend = "auto", n_threads = 4)
table(km$cluster)
km$parameters$tuning