Properly namespace the base oml library and make its archive correct.

.merlin

@@ -1,5 +1,8 @@
 PKG kaputt
-S ./src/lib/**
+PKG lacaml lbfgs ocephes
+S src/**
+S src-full/**
+S test
+S scripts
 B _build/**
 B _driver/**
-PKG lacaml lbfgs ocephes

src-full/cls/descriminant.ml

@@ -4,7 +4,7 @@ open Lacaml.D
 open Oml_util
 module LU = Lacaml_util
 
-module LDA(Data : Cls_intf.Continuous_encoded_data) = struct
+module LDA(Data : Oml_cls_intf.Continuous_encoded_data) = struct
   type feature = Data.feature
   type clas = Data.clas
 
@@ -100,7 +100,7 @@ module LDA(Data : Cls_intf.Continuous_encoded_data) = struct
 
 end (* LDA *)
 
-module QDA(Data : Cls_intf.Continuous_encoded_data) = struct
+module QDA(Data : Oml_cls_intf.Continuous_encoded_data) = struct
   type feature = Data.feature
   type clas = Data.clas
 

src-full/cls/descriminant.mli

@@ -5,17 +5,17 @@
     then used to model a {{:https://en.wikipedia.org/wiki/Multivariate_normal_distribution}
     Multivariate normal distribution}. These, per class, distributions are used
     in Bayes's rule for classification. *)
-module LDA(D: Cls_intf.Continuous_encoded_data) : sig
-  include Cls_intf.Generative with type feature = D.feature
+module LDA(D: Oml_cls_intf.Continuous_encoded_data) : sig
+  include Oml_cls_intf.Generative with type feature = D.feature
                           and type clas = D.clas
                           and type feature_probability = float
 
   val opt : ?shrinkage:float -> unit -> opt
 
 end
 
-module QDA(D: Cls_intf.Continuous_encoded_data) : sig
-  include Cls_intf.Generative with type feature = D.feature
+module QDA(D: Oml_cls_intf.Continuous_encoded_data) : sig
+  include Oml_cls_intf.Generative with type feature = D.feature
                           and type clas = D.clas
                           and type feature_probability = float
 

src-full/cls/logistic_regression.ml

@@ -16,7 +16,7 @@
 *)
 
 open Oml_util
-open Cls_intf
+open Oml_cls_intf
 
 module LrCommon(Data: Continuous_encoded_data) = struct
 

src-full/cls/logistic_regression.mli

@@ -30,9 +30,9 @@
   A constant [1] is added to all encoded features by [estimate],
   there is no need to add one with
   {{!val:Cls_intf.Continuous_encoded_data.encoding}encoding}. *)
-module Binary(D: Cls_intf.Continuous_encoded_data) :
+module Binary(D: Oml_cls_intf.Continuous_encoded_data) :
   sig
-    include Cls_intf.Classifier with  type feature = D.feature
+    include Oml_cls_intf.Classifier with  type feature = D.feature
                             and type clas = D.clas
 
     (** [opt ~lambda ~tolerance ()] a constructor for the optional arguments,
@@ -67,9 +67,9 @@ module Binary(D: Cls_intf.Continuous_encoded_data) :
   A constant [1] is added to all encoded features by [estimate],
   there is no need to add one with
   {{!val:Cls_intf.Continuous_encoded_data.encoding}encoding}. *)
-module Multiclass(D: Cls_intf.Continuous_encoded_data) :
+module Multiclass(D: Oml_cls_intf.Continuous_encoded_data) :
   sig
-    include Cls_intf.Classifier with (* type opt := opt
+    include Oml_cls_intf.Classifier with (* type opt := opt
                             and*) type feature = D.feature
                             and type clas = D.clas
 

src-full/cls/oml_naive_bayes.ml

@@ -18,10 +18,10 @@
 include Oml_lite_naive_bayes
 open Oml_util
 
-module Gaussian(Data: Cls_intf.Continuous_encoded_data) = struct
+module Gaussian(Data: Oml_cls_intf.Continuous_encoded_data) = struct
 
   module D = Distributions
-  module O = Online
+  module O = Oml_online
 
   type samples = (Data.clas * Data.feature) list
 
@@ -51,11 +51,11 @@ module Gaussian(Data: Cls_intf.Continuous_encoded_data) = struct
     let to_prior (prior, _) = prior in
     let to_likelihood (_, lkhd) =
       let indices = safe_encoding feature in
-      Common_naive_bayes.prod_arr2 (fun (mean,std) y ->
+      Oml_common_naive_bayes.prod_arr2 (fun (mean,std) y ->
         if std = 0. then 1. else D.normal_pdf ~mean ~std y)
         lkhd indices
     in
-    Common_naive_bayes.eval ~to_prior ~to_likelihood table
+    Oml_common_naive_bayes.eval ~to_prior ~to_likelihood table
 
   type opt = unit
   let default = ()
@@ -80,7 +80,7 @@ module Gaussian(Data: Cls_intf.Continuous_encoded_data) = struct
         let attr_params = Array.map select rsarr in
         (c, (class_prior, attr_params)))
     in
-    Common_naive_bayes.estimate "Gaussian"
+    Oml_common_naive_bayes.estimate "Gaussian"
       init update incorporate (module Cm)
 
 end (* Gaussian *)

src-full/cls/oml_naive_bayes.mli

@@ -22,8 +22,8 @@ include module type of Oml_lite_naive_bayes
   Gaussian Naive Bayes} by estimating mean and standard deviations
   for each of the quantitative features in the
   {{!modtype:Cls_intf.Continuous_encoded_data}encoded data}. *)
-module Gaussian(D: Cls_intf.Continuous_encoded_data) : sig
-  include Cls_intf.Generative with type feature := D.feature
+module Gaussian(D: Oml_cls_intf.Continuous_encoded_data) : sig
+  include Oml_cls_intf.Generative with type feature := D.feature
                               and type clas := D.clas
                               and type feature_probability = float array
 end

src-full/oml_full.ml

@@ -20,7 +20,7 @@ module Classification = struct
   module Naive_bayes = Oml_naive_bayes
   module Logistic_regression = Logistic_regression
   module Descriminant = Descriminant
-  module Performance = Performance
+  module Performance = Oml.Classification.Performance
 end
 
 module Regression = struct

src-full/rgr/oml_univariate.ml

@@ -15,7 +15,7 @@
    limitations under the License.
 *)
 
-include Common_univariate
+include Oml_common_univariate
 
 let confidence_interval, prediction_interval =
   let interval a lrm ~alpha x =
@@ -49,4 +49,3 @@ let beta_test ?(null=0.0) t =
 
 let coefficient_tests ?null t =
   [| alpha_test ?null t ; beta_test ?null t |]
-

src-full/stats/hypothesis_test.ml

@@ -16,7 +16,7 @@
 *)
 
 open Oml_util
-open Descriptive
+open Oml_descriptive
 open Distributions
 module F = Oml_functions
 

src-full/stats/hypothesis_test.mli

@@ -26,7 +26,7 @@
   we can still create a prediction interval based off of basic sampled
   statistics and Student's distribution.
   *)
-val prediction_interval : Descriptive.summary -> float -> float * float
+val prediction_interval : Oml_descriptive.summary -> float -> float * float
 
 (** A hypothesis test.  *)
 type t =
@@ -54,7 +54,7 @@ type null_hypothesis =
     hypothesis, where [d] is the difference between population parameter and
     the observed value, [e] is the standard error of the observed value, and
     [k] is the degrees of freedom in the statistical procedure.
-    
+
     One may think of this as a principled way to test the signal (diff)
     to noise (error) seen in a sample of data. *)
 val t_test : null_hypothesis -> degrees_of_freedom:int -> diff:float
@@ -81,5 +81,3 @@ val means_different_variance_test : null_hypothesis -> float array
 (** [variance_ratio_test sample1 sample2] tests the data in [sample1] and
     [sample2] have the same variance based on F-test.*)
 val variance_ratio_test : float array -> float array -> t
-
-

src-full/stats/oml_functions.ml

@@ -52,15 +52,16 @@ let rec regularized_beta ~alpha:a ~beta:b ?epsilon ?max_iterations =
                                 ((a +. (2. *. m)) *. (a +. (2. *. m) +. 1.0)) in
   let get_a _n _x = 1.0 in
   let log_beta = ln_beta a b in
-  let fraction = Continued_fraction.init ~get_a ~get_b in fun x ->
+  let fraction = Oml_continued_fraction.init ~get_a ~get_b in fun x ->
     if Oml_util.is_nan x || Oml_util.is_nan a || Oml_util.is_nan b ||
             x < 0.0 || x > 1.0 || a <= 0.0 || b <= 0.0 then nan
     else if (x > (a +. 1.) /. (2. +. b +. a) &&
                    1. -. x <= (b +. 1.) /. (2. +. b +. a))
     then 1. -. regularized_beta ~alpha:b ~beta:a ?epsilon ?max_iterations (1. -. x)
     else exp ((a *. log x) +. (b *. log1p (-.x)) -.
-                log a -. log_beta) *.
-                1.0 /. Continued_fraction.evaluate fraction ?epsilon ?max_iterations x
+              log a -. log_beta) *.
+         1.0 /.
+         Oml_continued_fraction.evaluate fraction ?epsilon ?max_iterations x
 
 let chi_square_less num_observations chi_square =
   regularized_lower_gamma ~a:((float num_observations) /. 2.0) (chi_square /. 2.0)

src/cls/cls_intf.ml → src/cls/oml_cls_intf.ml

@@ -95,7 +95,7 @@ module type Classifier = sig
 
   (** [eval classifier feature] assign {{!Probabilities.t}probabilities} to the
       possible {{!type:Data.clas}classes} based upon [feature]. *)
-  val eval : t -> feature -> clas Probabilities.t
+  val eval : t -> feature -> clas Oml_probabilities.t
 
   (** Representing training data. *)
   type samples = (clas * feature) list

src/cls/oml_lite_naive_bayes.ml

@@ -17,7 +17,7 @@
 
 open Oml_util
 
-module Binomial(Data: Cls_intf.Dummy_encoded_data) = struct
+module Binomial(Data: Oml_cls_intf.Dummy_encoded_data) = struct
 
   type feature = Data.feature
   type clas = Data.clas
@@ -51,16 +51,16 @@ module Binomial(Data: Cls_intf.Dummy_encoded_data) = struct
       let idx = safe_encoding b in
       if nb.e_bernoulli then
         let set = Array.to_list idx in
-        Common_naive_bayes.prod_arr (fun i ->
+        Oml_common_naive_bayes.prod_arr (fun i ->
           if List.mem i ~set then
             class_probs.(i)
           else
             (1.0 -. class_probs.(i)))
           (Array.init Data.size (fun x -> x))
       else
-        Common_naive_bayes.prod_arr (fun i -> class_probs.(i)) idx
+        Oml_common_naive_bayes.prod_arr (fun i -> class_probs.(i)) idx
     in
-    Common_naive_bayes.eval ~to_prior ~to_likelihood nb.table
+    Oml_common_naive_bayes.eval ~to_prior ~to_likelihood nb.table
 
   module Cm = Map.Make(struct type t = clas let compare = compare end)
 
@@ -75,7 +75,7 @@ module Binomial(Data: Cls_intf.Dummy_encoded_data) = struct
       arr
     in
     let incorporate all num_classes totalf =
-      let to_prob = Common_naive_bayes.smoothing_to_prob opt.smoothing in
+      let to_prob = Oml_common_naive_bayes.smoothing_to_prob opt.smoothing in
       List.map all ~f:(fun (cl, attr_count) ->
         let prior_count = float attr_count.(Data.size) in
         let likelihood =
@@ -88,7 +88,7 @@ module Binomial(Data: Cls_intf.Dummy_encoded_data) = struct
         cl, likelihood)
     in
     let table =
-      Common_naive_bayes.estimate "Binomial"
+      Oml_common_naive_bayes.estimate "Binomial"
         init update incorporate (module Cm) ?classes data
     in
     {table ; e_bernoulli = opt.bernoulli}
@@ -103,7 +103,7 @@ module Binomial(Data: Cls_intf.Dummy_encoded_data) = struct
 
 end (* Binomial *)
 
-module Categorical(Data: Cls_intf.Category_encoded_data) = struct
+module Categorical(Data: Oml_cls_intf.Category_encoded_data) = struct
 
   type feature = Data.feature
   type clas = Data.clas
@@ -140,9 +140,10 @@ module Categorical(Data: Cls_intf.Category_encoded_data) = struct
     let to_prior (prior, _) = prior in
     let to_likelihood (_, ftr_prob) =
       let indices = safe_encoding feature in
-      Common_naive_bayes.prod_arr2 (fun i lk_arr -> lk_arr.(i)) indices ftr_prob
+      Oml_common_naive_bayes.prod_arr2
+        (fun i lk_arr -> lk_arr.(i)) indices ftr_prob
     in
-    Common_naive_bayes.eval ~to_prior ~to_likelihood table
+    Oml_common_naive_bayes.eval ~to_prior ~to_likelihood table
 
   module Cm = Map.Make(struct type t = clas let compare = compare end)
 
@@ -154,7 +155,7 @@ module Categorical(Data: Cls_intf.Category_encoded_data) = struct
       (c + 1, arr)
     in
     let incorporate all num_classes totalf =
-      let to_prob = Common_naive_bayes.smoothing_to_prob opt in
+      let to_prob = Oml_common_naive_bayes.smoothing_to_prob opt in
       List.map all ~f:(fun (cl, (class_count, attr_count)) ->
         let prior      = to_prob (float class_count) totalf (float num_classes) in
         let likelihood =
@@ -167,7 +168,7 @@ module Categorical(Data: Cls_intf.Category_encoded_data) = struct
         in
         cl, (prior, likelihood))
     in
-    Common_naive_bayes.estimate "Categorical" init update incorporate
+    Oml_common_naive_bayes.estimate "Categorical" init update incorporate
       (module Cm)
 
   end (* Categorical *)

src/cls/oml_lite_naive_bayes.mli

@@ -19,8 +19,8 @@
   {{:https://en.wikipedia.org/wiki/Naive_Bayes_classifier}Naive Bayes}
   classifier on data encoded using
   {{!modtype:Cls_intf.Dummy_encoded_data}Dummy variables.} *)
-module Binomial(D: Cls_intf.Dummy_encoded_data) : sig
-  include Cls_intf.Generative with type feature = D.feature
+module Binomial(D: Oml_cls_intf.Dummy_encoded_data) : sig
+  include Oml_cls_intf.Generative with type feature = D.feature
                               and type clas = D.clas
                               and type feature_probability = float array
 
@@ -45,8 +45,8 @@ end
   {{:https://en.wikipedia.org/wiki/Naive_Bayes_classifier}Naive Bayes}
   classifier on data encoded using
   {{!modtype:Cls_intf.Category_encoded_data}Categorical variables.} *)
-module Categorical(D: Cls_intf.Category_encoded_data) : sig
-  include Cls_intf.Generative with type feature = D.feature
+module Categorical(D: Oml_cls_intf.Category_encoded_data) : sig
+  include Oml_cls_intf.Generative with type feature = D.feature
                               and type clas = D.clas
                               and type feature_probability = float array
 

src/oml.ml

@@ -3,24 +3,24 @@ module Util = Oml_util
 module Uncategorized = Oml_lite_uncategorized
 
 module Statistics = struct
-  module Continued_fraction = Continued_fraction
+  module Continued_fraction = Oml_continued_fraction
   module Functions = Oml_lite_functions
-  module Sampling = Sampling
-  module Descriptive = Descriptive
-  module Measures = Measures
+  module Sampling = Oml_sampling
+  module Descriptive = Oml_descriptive
+  module Measures = Oml_measures
 end
 
-module Online = Online
+module Online = Oml_online
 
 module Classification = struct
-  module Intf = Cls_intf
-  module Probabilities = Probabilities
+  module Intf = Oml_cls_intf
+  module Probabilities = Oml_probabilities
   module Naive_bayes = Oml_lite_naive_bayes
-  module Performance = Performance
+  module Performance = Oml_performance
 end
 
 module Regression = struct
   module Intf = Oml_lite_intf
   module Univariate = Oml_lite_univariate
-  module Interpolate = Interpolate
+  module Interpolate = Oml_interpolate
 end

src/oml.mli

@@ -12,29 +12,31 @@ end
 
 (** Inference, parameter estimation.*)
 module Statistics : sig
-  module Continued_fraction : sig include module type of Continued_fraction end
+  module Continued_fraction : sig include
+    module type of Oml_continued_fraction
+  end
   module Functions : sig include module type of Oml_lite_functions end
-  module Sampling : sig include module type of Sampling end
-  module Descriptive : sig include module type of Descriptive end
-  module Measures : sig include module type of Measures end
+  module Sampling : sig include module type of Oml_sampling end
+  module Descriptive : sig include module type of Oml_descriptive end
+  module Measures : sig include module type of Oml_measures end
 end
 
 (** Compute running statitics using recurrence equations. *)
-module Online : sig include module type of Online end
+module Online : sig include module type of Oml_online end
 
 (** Classify data based on features. *)
 module Classification : sig
-  module Intf : sig include module type of Cls_intf end
-  module Probabilities : sig include module type of Probabilities end
+  module Intf : sig include module type of Oml_cls_intf end
+  module Probabilities : sig include module type of Oml_probabilities end
   module Naive_bayes : sig
     include module type of Oml_lite_naive_bayes
   end
-  module Performance : sig include module type of Performance end
+  module Performance : sig include module type of Oml_performance end
 end
 
 (** Model relationship between variables. *)
 module Regression : sig
   module Intf : sig include module type of Oml_lite_intf end
   module Univariate : sig include module type of Oml_lite_univariate end
-  module Interpolate : sig include module type of Interpolate end
+  module Interpolate : sig include module type of Oml_interpolate end
 end