rename optimization rules file, add docstrings to averagers

src/AdvancedVI.jl

@@ -183,7 +183,7 @@ include("families/location_scale.jl")
 
 # Optimization Update Rules
 
-include("optimization/optimizers.jl")
+include("optimization/optrules.jl")
 
 export DoWG, DoG, COCOB
 

src/optimization/averaging.jl

@@ -1,4 +1,9 @@
 
+"""
+    NoAveraging()
+
+No averaging. This returns the last-iterate of the optimization rule.
+"""
 struct NoAveraging <: AbstractAverager end
 
 init(::NoAveraging, x) = x
@@ -7,6 +12,24 @@ apply(::NoAveraging, state, x) = x
 
 value(::NoAveraging, state) = state
 
+"""
+    PolynomialAveraging(eta)
+
+Polynomial averaging rule proposed [Shamir and Zhang](https://proceedings.mlr.press/v28/shamir13.html)
+At iteration `t`, the polynomial averaging rule is given as
+```julia
+    params_avg = (1 - w) * params_avg + w * x
+```
+where the averaging weight is 
+```julia
+    w = (eta + 1) / (t + eta)
+```
+Higher `eta` down-weights earlier iterations.
+The [DoG paper](https://arxiv.org/abs/2302.12022) suggests `eta = 8`.
+
+# Parameters
+- eta: Regularization term. (default: `8`)
+"""
 struct PolynomialAveraging{F} <: AbstractAverager 
     eta::F
 end