Added bibliography references

3 files changed, 158 insertions, 30 deletions

diff --git a/scilab_doc/neldermead/neldermead-bibliography-so.pdf b/scilab_doc/neldermead/neldermead-bibliography-so.pdf
new file mode 100644
index 0000000..33e0fdb
--- /dev/null
+++ b/scilab_doc/neldermead/neldermead-bibliography-so.pdf
diff --git a/scilab_doc/neldermead/neldermead.bib b/scilab_doc/neldermead/neldermead.bib
index 6f0ed09..3d6f69b 100644
--- a/scilab_doc/neldermead/neldermead.bib
+++ b/scilab_doc/neldermead/neldermead.bib
@@ -9,7 +9,6 @@ volume = {9},
 number = {1},
 pages = {112-147},
 keywords = {direct search methods; Nelder--Mead simplex methods; nonderivative optimization},
-note = {\url{http://link.aip.org/link/?SJE/9/112/1}},
 doi = {10.1137/S1052623496303470}
 }
 
@@ -19,7 +18,6 @@ doi = {10.1137/S1052623496303470}
    abstract = {A method is described for the minimization of a function of n variables, which depends on the comparison of function values at the (n + 1) vertices of a general simplex, followed by the replacement of the vertex with the highest value by another point. The simplex adapts itself to the local landscape, and contracts on to the final minimum. The method is shown to be effective and computationally compact. A procedure is given for the estimation of the Hessian matrix in the neighbourhood of the minimum, needed in statistical estimation problems. 10.1093/comjnl/7.4.308},
    author = {Nelder, J. A.  and Mead, R. },
    citeulike-article-id = {3009487},
-   doi = {http://dx.doi.org/10.1093/comjnl/7.4.308},
    journal = {The Computer Journal},
    month = {January},
    number = {4},
@@ -40,7 +38,6 @@ pages = { ix, 227 p. : },
 year = { 1979 },
 type = { Book },
 subjects = { Numerical analysis -- Data processing },
-catalogue-url = { http://nla.gov.au/nla.cat-vn1060620 },
 }
 
 @Book{Kelley1999,
@@ -49,14 +46,12 @@ author = {C. T. Kelley},
 publisher = {SIAM Frontiers in Applied Mathematics},
 volume={19},
 year={1999},
-note={\url{http://www.siam.org/books/textbooks/fr18_book.pdf}}
 }
 
 @Book{KelleyMethodsOptimizationMatlabCodes,
 title={Iterative Methods for Optimization: Matlab Codes},
 author = {C. T. Kelley},
 publisher = {North Carolina State University},
-note={\url{http://www4.ncsu.edu/~ctk/matlab_darts.html}}
 }
  
     @article{Spendley1962,
@@ -69,30 +64,13 @@ note={\url{http://www4.ncsu.edu/~ctk/matlab_darts.html}}
      number = {4},
      jstor_formatteddate = {Nov., 1962},
      pages = {441--461},
-     url = {http://www.jstor.org/stable/1266283},
      ISSN = {00401706},
      abstract = {A technique for empirical optimisation is presented in which a sequence of experimental designs each in the form of a regular or irregular simplex is used, each simplex having all vertices but one in common with the preceding simplex, and being completed by one new point. Reasons for the choice of design are outlined, and a formal procedure given. The performance of the technique in the presence and absence of error is studied and it is shown (a) that in the presence of error the rate of advance is inversely proportional to the error standard deviation, so that replication of observations is not beneficial, and (b) that the "efficiency" of the technique appears to increase in direct proportion to the number of factors investigated. It is also noted that, since the direction of movement from each simplex is dependent solely on the ranking of the observations, the technique may be used even in circumstances when a response cannot be quantitatively assessed. Attention is drawn to the ease with which second-order designs having the minimum number of experimental points may be derived from a regular simplex, and a fitting procedure which avoids a direct matrix inversion is suggested. In a brief appendix one or two new rotatable designs derivable from a simplex are noted.},
      publisher = {American Statistical Association and American Society for Quality},
      language = {},
-     copyright = {Copyright © 1962 American Statistical Association and American Society for Quality},
+     copyright = {Copyright 1962 American Statistical Association and American Society for Quality},
      year = {1962},
     }
-@book{citeulike:2815920,
-        abstract = {The only book on the market devoted to sequential simplex optimization This book presents an easy-to-learn, effective optimization technique that can be applied immediately to many problems in the real world. The sequential simplex is an evolutionary operation (EVOP) technique that uses experimental results- it does not require a mathematical model. The authors present their subject with a level of detail and clarity that is refreshingly welcome in a technical text. The basics are presented first, followed by a detailed discussion of the fine points needed to get the most out of this optimization technique. Worksheets are provided and their use is illustrated with step-by-step worked examples. This makes the logic and calculations of the simplex algorithms easy to understand and follow. The text also provides more than 200 figures and over 500 references to sequential simplex applications, which allows rapid access to specific examples of the use of the technique in a wide range of applications. Sequential Simplex Optimization: A Technique for Improving Quality and Productivity in Research, Development, and Manufacturing is essential for any student or professional who desires to learn this innovative technique quickly and easily.},
-        author = {Walters, Fred  H.  and Jr, Lloyd  R.  and Morgan, Stephen  L.  and Deming, Stanley  N. },
-        citeulike-article-id = {2815920},
-        howpublished = {Hardcover},
-        isbn = {0849358949},
-        keywords = {multiobjective, optimization, simplex},
-        month = {August},
-        posted-at = {2008-10-23 13:13:43},
-        priority = {2},
-        publisher = {CRC},
-        series = {Chemometrics series},
-        title = {Sequential Simplex Optimization: A Technique for Improving Quality and Productivity in Research, Development, and Manufacturing},
-        url = {http://www.amazon.ca/exec/obidos/redirect?tag=citeulike09-20\&path=ASIN/0849358949},
-        year = {1991}
-}
 @book{NumericalRecipes,
 author = {W. H. Press and Saul A. Teukolsky and William T. Vetterling and Brian P. Flannery},
 title = {Numerical Recipes in C, Second Edition},
@@ -253,8 +231,7 @@ publisher= {}}
     title = {Multiobjective Optimization in Engineering Design: Application to Fluid Power Systems},
     institution = {Department of Mechanical Engineering, Link\"oping University},
     year = {2001},
-  note={\url{https://polopoly.liu.se/content/1/c6/10/99/74/phdthesis.pdf}, 
-  \url{http://www.iei.liu.se/machine/johan-olvander}}
+  note={\url{https://polopoly.liu.se/content/1/c6/10/99/74/phdthesis.pdf}}
 }
 
 @article{Box1965,
@@ -487,3 +464,30 @@ PAGES  = {2928}
     pages = {396--401}
 }
 
+@book{WaltersParkerMorganDeming1991,
+ author = {Walters, F. S. and  Parker, L. R. and  Morgan, S. L. and Deming, S. N.},
+ title = {Sequential Simplex Optimization for Quality and Productivity in Research, Development, and Manufacturing},
+ year = {1991},
+ publisher= {CRC Press, Boca Raton, FL},
+ series = {Chemometrics series},
+}
+
+@ARTICLE{VanDerWiel1980,
+  AUTHOR = {Van Der Wiel, P.F.A. },
+  TITLE  = {Improvement of the super modified simplex optimisation procedure},
+  YEAR  = {1980},
+  JOURNAL  = {Analytica Chemica Acta},
+  VOLUME  = {122},
+  PAGES  = {421?-433},
+}
+
+@ARTICLE{Routh1977,
+    author = {Routh, M. W. and Swartz, P.A. and Denton, M.B.},
+    title = {Performance of the Super Modified Simplex},
+    journal = {Analytical Chemistry},
+    year = {1977},
+    volume = {49},
+    number = {9},
+    pages = {1422--1428}
+}
+
diff --git a/scilab_doc/neldermead/nmbibliography.tex b/scilab_doc/neldermead/nmbibliography.tex
index 6b525c0..8aee50c 100644
--- a/scilab_doc/neldermead/nmbibliography.tex
+++ b/scilab_doc/neldermead/nmbibliography.tex
@@ -1,7 +1,7 @@
 \chapter{Nelder-Mead bibliography}
 
 In this section, we present a brief overview of selected 
-papers, sorted in chronological order, which deals with 
+papers, sorted in chronological order, which deal with 
 the Nelder-Mead algorithm
 
 \section{Spendley, Hext, Himsworth, 1962}
@@ -98,7 +98,7 @@ The initial simplex is computed axis-by-axis, given the initial guess
 and a vector of step lengths.
 A factorial test is used to check if the computed optimum point is a local minimum.
 
-\subsection{Parkinson and Hutchinson, 1972}
+\section{Parkinson and Hutchinson, 1972}
 
 In \cite{parkinson1972}, "An investigation into the efficiency of variants on the simplex method", 
 Parkinson and Hutchinson explored 
@@ -145,6 +145,131 @@ complex points, and every point on the line segment joining these two points
 all have functional values lower than the functional values at each of the
 remaining complex points.
 
+\section{Routh, Swartz, Denton, 1977}
+
+"Performance of the Super-Modified Simplex",
+M.W. Routh, P.A. Swartz, M.B. Denton,
+Analytical Chemistry,
+1977
+
+In this article \cite{Routh1977}, 
+Routh, Swartz and Denton present a variant
+of the Nelder-Mead algorithm, which is called
+the Modified Simplex Method (SMS) in their
+paper. The algorithm is modified in the following
+way. After determination of the worst response (W), 
+the responses at the centroid (C) and reflected (R)
+vertices are measured and a second-order polynomial
+curve is fitted to the responses at W, C and R.
+Furthermore, the curve is extrapolated beyond W 
+and R by a percentage of the W-R vector resulting
+in two types of curve shapes. In the concave down
+case, a maximum occurs within the interval.
+Assuming a maximization process, evaluation of the
+derivative of the curve reveals the location of the
+predicted optimum whose response is subsequently
+evaluated, the new vertex is located at that position,
+and the optimization process is continued.
+In the concave up case, a response maximum
+does not occur within the interval so the extended
+interval boundary producing the highest predicted
+response is chosen as the new vertex location,
+its response is determined, and the optimization
+is continued. If the response at the predicted
+extended interval boundary location does not
+prove to be greater than the response at R, the 
+vertex R may instead be retained as the new 
+vertex and the process continued.
+The slope at the extended interval boundary
+may additionally be evaluated dictating the 
+magnitude of the expansion coefficient, i.e.
+the greater the slope (indicating rapid approach
+to the optimum location), the smaller the required
+expansion coefficient and, conversely, the smaller
+the slope (indicating remoteness from the 
+optimum location), the larger the required expansion
+coefficient.
+
+Some additional safeguard procedure must be used
+in order to prevent the collapse of the simplex.
+
+\section{Van Der Wiel, 1980}
+
+"Improvement of the Super-Modified Simplex
+Optimization Procedure",
+P.F.A., Van Der Wiel
+Analytica Chimica Acta,
+1980
+
+In this article \cite{VanDerWiel1980}, Van Der Wiel
+tries to improve the SMS method by Routh et al..
+His modifications are based on a Gaussian fit,
+weighted reflection point and estimation of response
+at the reflection point. Van Der Wiel presents 
+a simplified pseudo-code for one algorithm
+The method is tested in 5 cases, where the cost
+function is depending on the exponential function.
+
+\section{Walters, Parker, Morgan and Deming, 1991}
+
+"Sequential Simplex Optimization for Quality and Productivity in Research, Development, and Manufacturing", 
+F. S. Walters, L. R. Parker, Jr., S. L. Morgan, and S. N. Deming, 1991
+
+In this book \cite{WaltersParkerMorganDeming1991}, Walters, Parker, Morgan and Deming
+give a broad view on the simplex methods in chemistry. 
+The Spendley et al. and Nelder-Mead algorithms are particularily deeply analyzed, 
+with many experiments analyzed in great detail.
+Template tables are given, so that an engineer can manually perform the optimization
+and make the necessary calculations.
+Practical advices are given, which allow to make a better use of the algorithms.
+
+In chapter 5, "Comments on Fixed-size and Variable-size Simplexes",
+comparing the path of the two algorithms allows to check that a real optimum
+has been found.
+When the authors analyze the graph produced by the response depending on the number of 
+iteration, the general behavior of the fixed-size algorithm is made of four steps.
+Gains in response are initially rapid, but the rate of return decreases as the 
+simplex probes to find the ridge and then moves along the shallower 
+ridge to find the optimum.
+The behavior from different starting locations is also analyzed.
+Varying the size of the initial simplex is also analyzed for the fixed-size 
+simplex algorithm. The many iterations which are produced when a tiny 
+initial simplex is used with the fixed-size simplex is emphasized.
+
+The chapter 6, "General Considerations", warns that the user may 
+setup an degenerate initial simplex, leading to a false convergence 
+of the algorithm. Various other initial simplices are analyzed.
+Modifications in the algorithm to take into account for bounds 
+contraints are presented.
+The behavior of the fixed-size and variable-size simplex algorithms 
+is analyzed when the simplex converges.
+The "k+1" rule, introduced by Spendley et al. to take into account 
+for noise in the cost function is presented.
+
+The chapter 7, "Additional Concerns and Topics" deals with advanced questions 
+regarding these algorithms. The variable size simplex algorithm is analyzed
+in the situation of a ridge. Partially oscillatory collapse of the 
+Nelder-Mead algorithm is presented. The same behavior is 
+presented in the case of a saddle point. This clearly shows that 
+practionners were aware of the convergence problem of this 
+algorithm well before Mc Kinnon presented a simple counter example (in 1998).
+The "Massive Contraction" step of Nelder and Mead is presented as a 
+solution for this oscillatory behavior. The authors present 
+a method, due to Ernst, which allows to keep the volume of the 
+simplex, instead of shrinking it. This method is based on a 
+translation of the simplex. This modification requires 
+$n+1$ function evaluations. A more efficient method, due to King,
+is based on reflection with respect to the next-to-worst 
+vertex. This modification was first suggested by Spendley et al.
+in their fixed-size simplex algorithm.
+
+In the same chapter, the authors present the behavior of the algorithms 
+in the case of multiple optima. They also present briefly other 
+types of simplex algorithms.
+
+A complete bibliography (from 1962 to 1990) on simplex-based optimization is given in 
+the end of the book.
+
 \section{Subrahmanyam, 1989}
 
 "An extension of the simplex method to constrained nonlinear optimization",
@@ -235,7 +360,7 @@ accumulation point of the simplices is a critical point of f.
 
 The same ideas are presented in the book \cite{Kelley1999}.
 
-\subsection{Han, 2000}
+\section{Han, 2000}
 
 In his Phd thesis \cite{Han2000}, Lixing Han analyzes the 
 properties of the Nelder-Mead algorithm.
@@ -339,7 +464,7 @@ Johan Andersson, 2001
 This PhD thesis \cite{Andersson01multiobjectiveoptimization} gives a brief overview of the Complex method by Box in 
 section 5.1.
 
-\subsection{Peters, Bolte, Marschner, N\"{u}ssen and Laur, 2002}
+\section{Peters, Bolte, Marschner, N\"{u}ssen and Laur, 2002}
 
 In \cite{590075}, "Enhanced Optimization Algorithms for the Development of Microsystems", 
 the authors combine radial basis function interpolation methods 
@@ -396,4 +521,3 @@ Singer and Nelder
 This article is a complete review of the Nelder-Mead algorithm.
 Restarting the algorithm is adviced when a premature termination occurs.
 
-