1.5-8 (13 Mar 2023) ----- converted sprintf to snprintf 1.5-8 (22 Oct 2022) ----- solved clang-15 warning problems 1.5-7 (15 Jun 2022) ----- akima out again for interp, and now conditionally loading suggests 1.5-6 (02 Apr 2022) ----- akima is back and now USE_FC_LEN and DOUBLE_ for Brian other small documentation changes 1.5-5 (05 Sep 2019) ----- temporarily removed akima suggests 1.5-4 (27 Aug 2019) ----- copied over tgp's updated linalg.c/.h returning optimized value as out$value in alcoptGP and alcoptGPsep Simplified linalg.c/,h BLAS and Lapack headers, primarily for monomvn, but copied from tgp Fixed aGPsep bug with vectorized d$start argument Made sure all .C have PACKAGE argument Removed progress function from ALBO demos because $prog is returned added reset in mleGPsep when init is on boundary corrected imethod for alcray in aGPsep 1.5-3 (06 Dec 2018) ----- Fixed some method checkes in function headers Furong added readLines Fixed bug in IECI when Xref != Xcand Fixed end < close <= nrow(X) bug found by Mickael 1.5-2 (23 Jul 2018) ----- Fixed omp being used in examples so that they will all work on machines without OpenMP Documentation edits and nonug predGP and IECI option(s) Small documentation changes, and data.matrix now used instead of matrix to convert Xref 1.5-1 (15 Oct 2017) ----- Fixed NN bug in aGPsep 1.5 (22 Sep 2017) --- Alpha implementation of derivatives of ALC, along with n^2 economy in original ALC and IECI implementations added bootstrapped block latin hypercube functionality for consistent global lengthscale estimation 1.4 (31 May 2017) --- Added param = "both" option to mleGPsep for combined gradient MLE calculations to speed convergence dealt with init.c for registring entry points into C 1.3-2 (8 Jan 2017) ----- Turns out that strrep is too new, using paste/rep instead to solve R_C_BOUNDS_CHECK issue 1.3-1 (28 Dec 2016) ----- Added IECI capability (alpha functionality) Fixed R_C_BOUNDS_CHECK=yes problem pointed out by Kurt 1.3 (05 Aug 2016) --- updates for JSS article changed to VT contact information added mixed constraints demos from Picheny et al. paper and changed some optim.auglag defaults to match those used for the figures in the paper Fixed checking of end versus nrow(X) which was leading to a segfault in low nrow(X) examples; thanks to Jakob Richter for the bug report updates to optim.auglag for slack variables implementation and mixed equality and inequality constraints 1.2-1 (18 Sept 2015) ----- Fixed assert errors in gp_sep.cc for Windows build 1.2 (10 Sept 2015) --- vignette accepted at JSS -- includes latest modifications changed my to MY for myprintf, etc., for Brian initial C-version of optim via lbfgsb with OpenMP critical pragmas, accompanying laGPsep and aGPsep functions for local modeling with a separable correlation (beta) implementation of EIC for optimization (Schonlau, etal, 1998) alpha implementation of AL optimization by slack variables 1.1-5 (22 May 2015) ----- New vignette from JSS revisions swapped sum_of_columns for min_of_columns in multi-row Xref case for laGP Documentation enhancements and update for JSS/vignette resubmission Option to use separable GP model in optim.auglag. Option to model objective in optim.auglag rather than assuming a known scaled linear objective 1.1-4 (02 Dec 2014) ----- added predGPsep to predict with separable GPs corrected three abs to fabs in matrix.c corrected one more hard-code of 8 threads in vignette added vignette("laGP") to most seealso sections in Rd fixes to vignette for JSS pre-screening removed calc.eis because C-implementation was not in src; can use version from plgp package instead 1.1-3 (30 Oct 2014) ----- fixed assert(gp) in gp_sep.c added code to max-out omp.threads and gpu.threads at the OpenMP max in the system configuration changed vignette to default to two threads unless env specifies a different OMP_NUM_THREADS 1.1-2 (16 Oct 2014) ----- added separable GP (GPsep) for finding MLE separable lengthscale parameters for scaling isotropic versions ported calib functions over to package with documentation and examples added first pass at vignette 1.1-1 (02 Sept 2014) ----- avoid re-calulating derivative info when optimizing over the nugget moved several Suggests to Depends for R CHECK submitted rays paper and added to references now beta ray-based search including option to control the number of rays now outputting close in object to inspect fidelity of search 1.1 (20 March 2014) --- added ray-based search for faster greedy local design (alpha functionality) added functions for optimization under constraints by the augmented Lagrangian (alpha functionality) minor bug fixes and documentation updates removed snow; only supporting parallel now