Fix a few problems

src/fletchcr.jl

@@ -1,12 +1,4 @@
-#   Source:  The second problem given by
-#   R. Fletcher,
-#   "An optimal positive definite update for sparse Hessian matrices"
-#   Numerical Analysis report NA/145, University of Dundee, 1992.
-
-#   Scaled version.
-
-#   SIF input: Nick Gould, Oct 1992.
-
+#   Source:
 #   problem 32 in
 #   L. Luksan, C. Matonoha and J. Vlcek
 #   Modified CUTE problems for sparse unconstrained optimization,
@@ -16,6 +8,24 @@
 #
 #   http://www.cs.cas.cz/matonoha/download/V1081.pdf
 #
+#   CUTEst cites
+#     The second problem given by
+#     R. Fletcher,
+#     "An optimal positive definite update for sparse Hessian matrices"
+#     Numerical Analysis report NA/145, University of Dundee, 1992.
+#
+#     Scaled version.
+#
+#     SIF input: Nick Gould, Oct 1992.
+#
+#   as source for this problem.
+#   It is possible that Fletcher may have used this problem in his
+#   technical report, but the published version of his report
+#
+#     SIAM J. Optimization, 5(1), pp 192-218, 1995
+#
+#   uses the chained Rosenbrock problem, not the one below.
+#
 #   classification OUR2-AN-V-0
 #
 # J.-P. Dussault, Rennes 09/2015.

src/hs105.jl

@@ -63,7 +63,7 @@ function hs105(args...)
   @NLobjective(
     nlp,
     Min,
-    - sum(log(a[i] + b[i] + c[i])/sqrt(2*pi) for i=1:235)
+    - sum(log((a[i] + b[i] + c[i])/sqrt(2*pi)) for i=1:235)
   )
 
   return nlp

src/hs55.jl

@@ -22,7 +22,7 @@ function hs55(args...)
   @variable(nlp, 0 <= x[i=1:6] <= uvar[i], start = x0[i])
 
   @constraint(nlp, x[1] + 2*x[2] + 5*x[5] - 6 == 0)
-  @constraint(nlp, x[3] + x[2] + x[3] - 3 == 0)
+  @constraint(nlp, x[1] + x[2] + x[3] - 3 == 0)
   @constraint(nlp, x[4] + x[5] + x[6] - 2 == 0)
   @constraint(nlp, x[1] + x[4] - 1 == 0)
   @constraint(nlp, x[2] + x[5] - 2 == 0)
@@ -35,4 +35,4 @@ function hs55(args...)
   )
 
   return nlp
-end
+end

src/hs70.jl

@@ -32,10 +32,12 @@ function hs70(args...)
   yobs[11:19] = [.702,.528,.385,.257,.159,.0869,.0453,.01509,.00189]
 
   @NLexpression(nlp, b, x[3] + (1-x[3])*x[4])
-  @NLexpression(nlp, ycal[i=1:19], (1 + 1/(12*x[2]))*(x[3]*b^x[2])*((x[2]/6.2832)^(0.5))
-    *(c[i]/7.685)^(x[2] - 1)*exp(x[2] - b*c[i]*x[2]/7.658)
-    + (1 + (1/(12*x[1])))*(1 - x[3])*(b/x[4])^x[1]*(x[1]/6.2832)^0.5
-    * (c[i]/7.658)^(x[1]-1)*exp(x[1] - b*c[i]*x[1]/(7.658*x[4]))  
+  @NLexpression(nlp,
+                ycal[i=1:19],
+                (1 + 1 / (12 * x[2])) * (x[3] * b^x[2] * (x[2] / 6.2832)^(0.5) *
+                                         (c[i] / 7.685)^(x[2] - 1) * exp(x[2] - b * c[i] * x[2] / 7.658)) +
+                (1 + 1 / (12 * x[1])) * (1 - x[3]) * (b / x[4])^x[1] * (x[1] / 6.2832)^0.5 *
+                (c[i] / 7.658)^(x[1] - 1) * exp(x[1] - b * c[i] * x[1] / (7.658 * x[4]))
   )
 
   @NLconstraint(nlp, x[3] + (1-x[3])*x[4] >= 0)
@@ -47,4 +49,4 @@ function hs70(args...)
   )
 
   return nlp
-end
+end

src/hs93.jl

@@ -27,11 +27,11 @@ function hs93(args...)
   @NLobjective(
     nlp,
     Min,
-    0.204*x[1]*x[4]*(x[1] + x[2] + x[3]) 
+    0.0204*x[1]*x[4]*(x[1] + x[2] + x[3])
     + 0.0187*x[2]*x[3]*(x[1] + 1.57*x[2] + x[4])
     + 0.0607*x[1]*x[4]*x[5]^2*(x[1] + x[2] + x[3]) 
     + 0.0437*x[2]*x[3]*x[6]^2*(x[1] + 1.57*x[2] + x[4])
   )
 
   return nlp
-end
+end

test/compare_cutest.jl

@@ -41,11 +41,11 @@ function compute_status(probname; showvals=false)
         x0 = optnlp.meta.x0
         if !(obj(cutenlp, x0) ≈ obj(optnlp, x0))
             sameobjval = false
-            showvals && @show obj(cutenlp, x0) obj(d, x0)
+            showvals && @show obj(cutenlp, x0) obj(optnlp, x0)
         end
         if !(grad(cutenlp, x0) ≈ grad(optnlp, x0))
             samegradval = false
-            showvals && @show [grad(cutenlp, x0) grad(d, x0)]
+            showvals && @show [grad(cutenlp, x0) grad(optnlp, x0)]
         end
         isok = true
     end