Agile Combinatorial Factor Decomposition (AgileFD) Modified 12/01/2016 Version 1.2 ================================================ Copyright 2016 Institute for Computational Sustainability Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. ================================================ *** Main Contributors *** Junwen Bai (Cornell University Dept. of Computer Science) jb2467@cornell.edu Richard Bernstein (Cornell University Dept. of Computer Science) rab38@cornell.edu Johan Björck (Cornell University Dept. of Computer Science) njb225@cornell.edu Ronan Le Bras (Cornell University Dept. of Computer Science) lebras@cs.cornell.edu Brendan Rappazzo (Cornell University Dept. of Computer Science) bhr54@cornell.edu Yexiang Xue (Cornell University Dept. of Computer Science) yx247@cornell.edu *** Citation *** This work accompanies the following publications: Xue, Y., Bai, J., Le Bras, R., Rappazzo, B., Bernstein, R., Bjorck, J., Longpre, L., Suram, S., van Dover, B., Gregoire, J., Gomes, C (2017). Phase-Mapper: An AI Platform to Accelerate High Throughput Materials Discovery. In Twenty-ninth International Conference on Innovative Applications of Artificial Intelligence (IAAI'17). Suram, S.K., Xue, Y., Bai, J., LeBras, R., Rappazzo, B.H., Bernstein, R., Bjorck, J., Zhou, L., van Dover, R.B., Gomes, C.P. and Gregoire, J.M. (2016). Automated Phase Mapping with AgileFD and its Application to Light Absorber Discovery in the V-Mn-Nb Oxide System. ACS Combinatorial Science. DOI: 10.1021/acscombsci.6b00153 Please cite these or related works as appropriate in follow-up publications. *** Introduction and Prerequisites *** AgileFD has two external library dependencies which must be installed: 1) Armadillo, see: http://arma.sourceforge.net 2) ILOG/CPLEX Optimization Studio: https://www.ibm.com/developerworks/downloads/ws/ilogcplex/ AgileFD also uses TCLAP (http://tclap.sourceforge.net/) to process command-line arguments and Tino Kluge's spline implementation (http://kluge.in-chemnitz.de/opensource/spline/). These are header-only includes and are provided in the source distribution. *** Compilation *** An example Makefile for GNU make and g++ is included, but will need to be modified to reflect the installation location of CPLEX, and the BLAS library (like openblas) used with Armadillo. Additional changes will be needed for alternate compilers. This configuration has been used on a few different Linux distributions, but additional changes could be required for other systems. *** Usage *** agilefd [-h] [OPTIONS] --inst INSTANCE_FILENAME --m M --k K --sol SOLUTION_FILENAME Options: --length The customized length of the array of qvalues (could be different from the length of the given qvalues in the instance file). This value overrides --stepsize if both are specified. --qmax The customized maximum q value (could be different from the maximum qvalue in the instance file) --qmin The customized minimum q value (could be different from the minimum qvalue in the instance file) --snapshot Every ? iterations, output one snapshot of the lastest solution --gibbs whether or not to enforce Gibbs phase rule --mipgap mipgap for MIP done via CPLEX, the default is 0.1 --sparsity The overall sparsity coefficient --stepsize Customized stepsize for resampling. A default stepsize of 0 will result in use of the standard stepsize. If both are specified, --length will take precedence. --sampleInit The filename containing initialzation from single-phase sample points --valueInit The filename containing initialization values for phases and phase freezing --seed Random seed for the random number generator. The default value of -1 means the random seed will be time(0)). --c Related to termination criterion: In one iteration, if (old_cost-new_cost) The maximum time(seconds) allowed to train the model --m (required) The number of possible different shifts --k (required) The number of phases --sol (required) The output filename of the solution --inst (required) Input instance file --, --ignore_rest Ignores the rest of the labeled arguments following this flag. --version Displays version information and exits. -h, --help Displays usage information and exits. *** Example *** agilefd --inst input/my_inst.txt --m 10 --k 8 --sol output/my_output.txt --time 1000 --c 1e-5 --sparsity 1.0 --gibbs --mipgap 0.1 --valueInit input/my_valueInit.txt --snapshot 300 --qmin 11.0 --qmax 59.0 --length 2000 The formats of the input instance file and the output solution file can be found in the following paper: Le Bras, R., Bernstein, R., Gregoire, J. M., Suram, S. K., Gomes, C. P., Selman, B., & van Dover, R. B. (2014). A Computational Challenge Problem in Materials Discovery: Synthetic Problem Generator and Real-World Datasets. In Twenty-Eighth International Conference on Artificial Intelligence (AAAI 14). [NOTE additions to solution file specific to AgileFD - H,HS,L] *** Value Initialization File Format *** A value initialization file specifies the basis patterns to use for initialization, as well as related configuration options. The format is as follows: // Comments can be included as complete lines beginning with two slashes // Q values corresponding to the basis vectors, in the same units as in the instance file (e.g. nm^-1 or A^-1). // Basis vectors are resampled, so these values do not need to exactly match the ones in the instance file. Q=1,1.1,1.2,... // Basis patterns are only shifted to the right (positive shift) in AgileFD, // so initial or frozen basis patterns should be specified as far // to the left as expected. The V parameter is a multiplicative shift which // is applied to the Q vector and affects all basis patterns in the same way. // The following effectively shifts all basis patterns 1% to the left. V=0.99 // B1...BK specify the basis patterns you wish to initialize. The indices must be // sequential, but fewer than K can be specified if desired. B1=0.3123,0.545234,...... B2=0.4324,0.454243,...... B3=0.42345,1.42344,...... ... // Whether to freeze each basis pattern: 0=no, 1=yes F1=1 F2=0 F3=1 ... // Lists sample indices in which each phase is allowed to appear. For example, // phase 1 could appear at sample points 1,2,54,65,76......, etc. If S is not // specified for an initialized basis, it can appear at any sample by default. S1=1,2,54,65,76,...... S2=1,2,4,7,111,...... S3=67,89,111,...... ... *** Sample Initialization File Format *** Sample initialization is an alternative to value initialization, where the desired basis patterns are taken from samples in the instance file. To initialize phase 1 using sample point 123 and phase 2 from sample point 56: B1=123 B2=56 Q should not be specified because the patterns come from the instance file. However, V, F, and S are set the same as in a value initialization file. *** Contact me *** If you have any questions or suggestions, please feel free to contact me. This is my email address: jb2467@cornell.edu.