Merge pull request #1 from PumasAI-Labs/juanjose/indent-fix

Change indent to 4

.JuliaFormatter.toml

@@ -1,2 +1,2 @@
-indent = 2
+indent = 4
 format_docstrings = true

01-linear_regression.jl

@@ -24,8 +24,8 @@ x = rand(uniform, 1, num_samples)  # samples stored columnwise
 y = true_function.(x) + σ * ϵ
 
 fig = scatter(vec(x), vec(y); axis = (xlabel = "x", ylabel = "y"), label = "data");
-lines!(-1..1, true_function; color = :gray, label = "true");
-axislegend(; position=:rb);
+lines!(-1 .. 1, true_function; color = :gray, label = "true");
+axislegend(; position = :rb);
 fig
 
 # 1.2. Model `true_function` with a linear regression model
@@ -41,6 +41,6 @@ ŷ = fitted_linreg(x)
 
 fig = scatter(vec(x), vec(y); axis = (xlabel = "x", ylabel = "y"), label = "data");
 scatter!(vec(x), vec(ŷ); label = "prediction");
-lines!(-1..1, true_function; color = :gray, label = "true");
-axislegend(; position=:rb);
+lines!(-1 .. 1, true_function; color = :gray, label = "true");
+axislegend(; position = :rb);
 fig

02-complex_relationships.jl

@@ -16,7 +16,7 @@ x = rand(uniform, 1, num_samples)
 y = true_function.(x) + σ * ϵ
 
 fig = scatter(vec(x), vec(y); axis = (xlabel = "x", ylabel = "y"), label = "data");
-lines!(-1..1, true_function; color = :gray, label = "true");
+lines!(-1 .. 1, true_function; color = :gray, label = "true");
 axislegend();
 fig
 
@@ -31,7 +31,7 @@ ŷ_ex22_50iter = fitted_linreg(x)
 
 fig = scatter(vec(x), vec(y); axis = (xlabel = "x", ylabel = "y"), label = "data");
 scatter!(vec(x), vec(ŷ_ex22_50iter); label = "prediction");
-lines!(-1..1, true_function; color = :gray, label = "true");
+lines!(-1 .. 1, true_function; color = :gray, label = "true");
 axislegend();
 fig
 
@@ -46,6 +46,6 @@ ŷ = fitted_nn(x)
 
 fig = scatter(vec(x), vec(y); axis = (xlabel = "x", ylabel = "y"), label = "data");
 scatter!(vec(x), vec(ŷ), label = "prediction");
-lines!(-1..1, true_function; color = :gray, label = "true");
+lines!(-1 .. 1, true_function; color = :gray, label = "true");
 axislegend();
 fig

03-bias-variance_tradeoff.jl

@@ -26,7 +26,7 @@ fig = scatter(vec(x), vec(y); axis = (xlabel = "x", ylabel = "y"), label = "data
 scatter!(vec(x), vec(ŷ_underfit), label = "prediction (5 iterations)");
 scatter!(vec(x), vec(ŷ), label = "prediction (50 iterations)");
 scatter!(vec(x), vec(ŷ_overfit), label = "prediction (1000 iterations)");
-lines!(-1..1, true_function; color = :gray, label = "true");
+lines!(-1 .. 1, true_function; color = :gray, label = "true");
 axislegend();
 fig
 
@@ -46,19 +46,15 @@ ŷ_linreg = fitted_linreg(x)
 fig = scatter(vec(x), vec(y); axis = (xlabel = "x", ylabel = "y"), label = "data");
 scatter!(vec(x), vec(ŷ_linreg), label = "$max_iterations iterations");
 scatter!(vec(x), vec(ŷ_ex22_50iter), label = "50 iterations");
-lines!(-1..1, true_function; color = :gray, label = "true");
+lines!(-1 .. 1, true_function; color = :gray, label = "true");
 axislegend();
 fig
 
 # 3.3. The impact of the NN size
 
 nn = MLPDomain(1, (32, tanh), (32, tanh), (1, identity); bias = true)
-fitted_nn = fit(
-    nn, 
-    target; 
-    optim_alg = DeepPumas.BFGS(), 
-    optim_options = (; iterations = 1_000)
-)
+fitted_nn =
+    fit(nn, target; optim_alg = DeepPumas.BFGS(), optim_options = (; iterations = 1_000))
 
 ŷ = fitted_nn(x)
 

04-generalization.jl

@@ -20,22 +20,23 @@ x_valid = rand(uniform, 1, num_samples)
 y_valid = true_function.(x_valid) + σ * ϵ_valid
 target_valid = preprocess(x_valid, y_valid)
 
-fig = scatter(vec(x_train), vec(y_train); axis = (xlabel = "x", ylabel = "y"), label = "training data");
+fig = scatter(
+    vec(x_train),
+    vec(y_train);
+    axis = (xlabel = "x", ylabel = "y"),
+    label = "training data",
+);
 scatter!(vec(x_valid), vec(y_valid); label = "validation data");
-lines!(-1..1, true_function; color = :gray, label = "true");
+lines!(-1 .. 1, true_function; color = :gray, label = "true");
 axislegend();
 fig
 
 # 4.2. Validation loss as a proxy for generalization performance
 
 loss_train_l, loss_valid_l = [], []
 
-fitted_nn = fit(
-    nn,
-    target_train;
-    optim_alg = DeepPumas.BFGS(),
-    optim_options = (; iterations = 10),
-)
+fitted_nn =
+    fit(nn, target_train; optim_alg = DeepPumas.BFGS(), optim_options = (; iterations = 10))
 push!(loss_train_l, sum((fitted_nn(x_train) .- y_train) .^ 2))
 push!(loss_valid_l, sum((fitted_nn(x_valid) .- y_valid) .^ 2))
 
@@ -112,6 +113,6 @@ ŷ_ho = nn_ho(x_valid)
 
 fig = scatter(vec(x_valid), vec(y_valid); label = "validation data");
 scatter!(vec(x_valid), vec(ŷ_ho), label = "prediction (hyperparam opt.)");
-lines!(-1..1, true_function; color = :gray, label = "true");
-axislegend(; position=:ct);
+lines!(-1 .. 1, true_function; color = :gray, label = "true");
+axislegend(; position = :ct);
 fig