; Standard k nearest neighbors classifier 
; Does not handle 
;   > 2 classes
;   even values of k
;   any bad input, really
;
; David Andrzejewski (andrzeje@cs.wisc.edu)
;
; Command-line arguments
; 0 training data filename 
; 1 test data filename
; 2 value of k (how many neighbors to use)
;
; EX: clj knn.clj train.txt test.txt 3
;
; Data file format is one example per line, with
; space-separated vector values. By convention, the 
; first element of each vector is the class label.
; 
(use '[clojure.contrib.duck-streams :only (read-lines)])
(use '[clojure.contrib.math :only (sqrt expt)])
(use '[clojure.contrib.seq-utils :only (indexed)])

(defstruct example :label :data)

(defn euclidean-distance
  "L2 distance between 2 vectors"
  [v1 v2]
  (sqrt (reduce + (map #(expt (- %1 %2) 2) v1 v2))))

(defn parse-line
  "parse a line of text-separated numbers"
  [line]
  (map #(Float/parseFloat %1) (.split line " ")))

(defn read-vector-file 
  "one space-separated vector per line, where the 
  first element of each vector is the label"
  [filename] 
  (for [line (read-lines filename)]
    (#(struct example (first %1) (rest %1)) (parse-line line))))

(defn value-count 
  "how many times does val occur in values?"
  [values val]
  (count (filter (partial == val) values)))

(defn majority-label 
  "what is the majority label among this set of examples?"
  [examples]
  (let [labels (map :label examples)]
    (last (sort-by (partial value-count labels) (set labels)))))

(defn nearest-neighbors
  "return the k nearest neighbors"
  [example dataset k]
  (take k (sort-by #(euclidean-distance (:data example) 
					(:data %1)) dataset)))

(defn classify-test 
  "classify each testset example"
  [trainset testset k]
  (for [testex testset]
    (majority-label (nearest-neighbors testex trainset k))))

(defn distance-matrix 
  "all pairwise distances as a vector of vectors"
  [vectors]
  (vec (for [v vectors] (map (partial euclidean-distance v) vectors))))


(let [trainfile (nth *command-line-args* 0)
      testfile (nth *command-line-args* 1)
      k (Integer/parseInt (nth *command-line-args* 2))]
  (doseq [[i label] (indexed (classify-test (read-vector-file trainfile)
                                            (read-vector-file testfile) k))]
    (println (format "Test example %d classified as %s" i (int label)))))