ufffffffffffffffffffff

src/direct_mlj.jl

@@ -83,7 +83,7 @@
     n_input = size(X, 1)
     dims = size(y)
     if length(dims) == 1
         n_output = 1
     else
         n_output = dims[1]
     end
@@ -110,7 +110,7 @@
     chain = Chain(
         Dense(n_input, 20, relu),
         Dense(20, 20, relu),
-        #Dense(20, 20, relu),
+        Dense(20, 20, relu),
         Dense(20, n_output)
     )
 
@@ -140,7 +140,7 @@
     y, decode = y
 
     if (m.model === nothing)
-        @warn "Warning: no Flux model has been provided in the model. LaplaceRedux will use a standard MLP with 3 hidden layers with 20 neurons each and input and output layers compatible with the dataset."
+        @warn "Warning: no Flux model has been provided in the model. LaplaceRedux will use a standard MLP with 2 hidden layers with 20 neurons each and input and output layers compatible with the dataset."
         shape = dataset_shape(m, X, y)
 
         m.model = default_build(11, shape)
@@ -182,7 +182,7 @@
 
         # Print loss every 100 epochs if verbosity is 1 or more
         if verbosity >= 1 && epoch % 100 == 0
             println("Epoch $epoch: Loss: $loss_per_epoch ")
         end
     end
 
@@ -386,10 +386,10 @@
     for name in names
         if !(name in exceptions)
             if !_isdefined(m1, name)
                 !_isdefined(m2, name) || return false
             elseif _isdefined(m2, name)
                 if name in MLJBase.deep_properties(LaplaceRegressor)
                     _equal_to_depth_one(getproperty(m1, name), getproperty(m2, name)) ||
                         return false
                 else
                     (
@@ -404,7 +404,7 @@
                     ) || return false
                 end
             else
                 return false
             end
         end
     end
@@ -419,21 +419,21 @@
 
 function _equal_flux_chain(chain1::Flux.Chain, chain2::Flux.Chain)
     if length(chain1.layers) != length(chain2.layers)
         return false
     end
     params1 = Flux.params(chain1)
     params2 = Flux.params(chain2)
     if length(params1) != length(params2)
         return false
     end
     for (p1, p2) in zip(params1, params2)
         if !isequal(p1, p2)
             return false
         end
     end
     for (layer1, layer2) in zip(chain1.layers, chain2.layers)
         if typeof(layer1) != typeof(layer2)
             return false
         end
     end
     return true
@@ -605,7 +605,7 @@
 
 # Hyperparameters (format: name-type-default value-restrictions)
 
-- `model::Union{Flux.Chain,Nothing} = nothing`:                                                     Either nothing or a Flux model provided by the user and compatible with the dataset. In the former case, LaplaceRedux will use a standard MLP with 3 hidden layer with 20 neurons each.
+- `model::Union{Flux.Chain,Nothing} = nothing`:                                                     Either nothing or a Flux model provided by the user and compatible with the dataset. In the former case, LaplaceRedux will use a standard MLP with 2 hidden layer with 20 neurons each.
 
 - `flux_loss = Flux.Losses.logitcrossentropy` :                                                     a Flux loss function
 
@@ -743,7 +743,7 @@
 
 # Hyperparameters (format: name-type-default value-restrictions)
 
-- `model::Union{Flux.Chain,Nothing} = nothing`:                                                     Either nothing or a Flux model provided by the user and compatible with the dataset. In the former case, LaplaceRedux will use a standard MLP with 3 hidden layer with 20 neurons each.
+- `model::Union{Flux.Chain,Nothing} = nothing`:                                                     Either nothing or a Flux model provided by the user and compatible with the dataset. In the former case, LaplaceRedux will use a standard MLP with 2 hidden layer with 20 neurons each.
 - `flux_loss = Flux.Losses.logitcrossentropy` :                                                     a Flux loss function
 
 - `optimiser = Adam()`                                                                              a Flux optimiser