Merge branch 'algorithm' into predict

Author: ablaom

docs/src/algorithm_traits.jl

@@ -69,7 +69,7 @@ Every LearnAPI method that is not a trait and which is specifically implemented
 return value is a tuple of symbols from this list: $(join(FUNCTIONS, ", ")). To regenerate
 this list, do `LearnAPI.functions()`.
 
-See also [`LearnAPI.Model`](@ref).
+See also [`LearnAPI.Algorithm`](@ref).
 
 """
 functions(::Type) = ()
@@ -100,13 +100,13 @@ Here's a sample implementation for a supervised model where predictions are ordi
 values of the target variable:
 
 ```julia
-@trait MyNewModel predict_proxy = LearnAPI.TrueTarget()
+@trait MyNewAlgorithm predict_proxy = LearnAPI.TrueTarget()
 ```
 
 which is shorthand for
 
 ```julia
-LearnAPI.predict_proxy(::Type{<:MyNewModelType}) = LearnAPI.TrueTarget()
+LearnAPI.predict_proxy(::Type{<:MyNewAlgorithmType}) = LearnAPI.TrueTarget()
 ```
 
 """
@@ -132,13 +132,13 @@ The possible return values for this trait are: `LearnAPI.JointSampleable()`,
 Here's a sample implementation:
 
 ```julia
-@trait MyNewModel predict_joint_proxy = LearnAPI.JointDistribution()
+@trait MyNewAlgorithmType predict_joint_proxy = LearnAPI.JointDistribution()
 ```
 
 which is shorthand for
 
 ```julia
-LearnAPI.predict_joint_proxy(::Type{<:MyNewModelType}) = LearnAPI.JointDistribution()
+LearnAPI.predict_joint_proxy(::Type{<:MyNewAlgorithmType}) = LearnAPI.JointDistribution()
 ```
 
 """

src/algorithm_traits.jl

@@ -1,21 +1,19 @@
-abstract type MLType end
+abstract type LearnAPIType end
 
 """
-    LearnAPI.Model
+    LearnAPI.Algorithm
 
-An optional abstract type for models implementing LearnAPI.jl.
+An optional abstract type for algorithms implementing LearnAPI.jl.
 
-If `typeof(m) <: LearnAPI.Model`, then `m` is guaranteed to be a model in the LearnAPI
-sense.
+If `typeof(alg) <: LearnAPI.Algorithm`, then `alg` is guaranteed to be an ML/statistical
+algorithm in the strict LearnAPI sense.
 
-# New model implementations
+# New implementations
 
-While not a formal requirement, model types implementing the LearnAPI interface are
-encouraged to subtype `LearnAPI.Model`, unless it is disruptive to do so.
+While not a formal requirement, algorithm types implementing the LearnAPI.jl are
+encouraged to subtype `LearnAPI.Algorithm`, unless it is disruptive to do so.
 
 See also [`LearnAPI.functions`](@ref).
 
 """
-abstract type Model <: MLType end
-
-# See src/model_traits.jl for the `methods` trait definition.
+abstract type Algorithm <: LearnAPIType end