Refactor benchmark files by removing unused includes and updating nonsmooth term initialization

Author: MohamedLaghdafHABIBOULLAH

paper/examples/Benchmark.jl

@@ -16,9 +16,6 @@ using LaTeXStrings
 include("comparison-config.jl")
 using .ComparisonConfig: CFG, CFG2
 
-include("Bench-utils.jl")
-using .BenchUtils
-
 #############################
 # ===== Helper utils ====== #
 #############################
@@ -39,14 +36,6 @@ function ensure_qn(model, which::Symbol)
     error("Unknown QN: $which (expected :LBFGS or :LSR1)")
 end
 
-# Close a PrettyTables Markdown string by adding a bottom rule
-function close_markdown_table(table_str::AbstractString)
-    lines = split(String(table_str), '\n')
-    isempty(lines) && return table_str
-    sep = length(lines) >= 2 ? lines[2] : repeat("-", 10)
-    push!(lines, sep)
-    return join(lines, '\n')
-end
 
 #############################
 # ======= SVM bench ======= #

paper/examples/example1.jl

@@ -6,7 +6,7 @@ Random.seed!(1234)
 model, nls_model, _ = RegularizedProblems.svm_train_model()  # Build SVM model
 f = LSR1Model(model)                                         # L-SR1 Hessian approximation
 λ = 1.0                                                      # Regularization parameter
-h = RootNormLhalf(1.0)                                       # Nonsmooth term
+h = RootNormLhalf(λ)                                         # Nonsmooth term
 reg_nlp = RegularizedNLPModel(f, h)                          # Regularized problem
 solver = R2NSolver(reg_nlp)                                  # Choose solver
 stats  = RegularizedExecutionStats(reg_nlp)

paper/paper.md

@@ -144,7 +144,7 @@ Random.seed!(1234)
 model, nls_model, _ = RegularizedProblems.svm_train_model()  # Build SVM model
 f = LSR1Model(model)                                         # L-SR1 Hessian approximation
 λ = 1.0                                                      # Regularization parameter
-h = RootNormLhalf(1.0)                                       # Nonsmooth term
+h = RootNormLhalf(λ)                                       # Nonsmooth term
 reg_nlp = RegularizedNLPModel(f, h)                          # Regularized problem
 solver = R2NSolver(reg_nlp)                                  # Choose solver
 stats  = RegularizedExecutionStats(reg_nlp)