[mlpack-git] mlpack-2.x's head updated: Merge branch 'master' into mlpack-2.x (281fb8a)

[gitdub at mlpack.org]

Repository : https://github.com/mlpack/mlpack
Branch 'mlpack-2.x' now includes:

     ab68cf1 improve speed of SparseAutoencoder and make it more flexible
     dde2b78 fix example of sparse_autoencoder
     d165890 reduce the cost of construction and destruction of the layers
     ea2d0dc refine the example; the size of the basis should not too big
     f0ec539 refine the example
     a094a62 refine the example
     92b4e44 can use Greedy parameters to determine the Gradient should omit some repeated calculation or not
     b7cfe68 add test cases of modified SparseAutoEncoder
     af6b87e add test
     30266a0 change the name of test suite to avoid definition confliction
     a1d37c5 change boost test header to avoid build issue
     f16a615 prevent nan/inf values
     c38a631 1 : Enhance SparseAutoencoder a : add functions Train b : add new constructor c : add Serialize
     0b03cdc update header and type define
     88d486a 1 : fix bug--Forward function of DropOutLayer cannot deduce correct type (because submat of arma is a proxy object) 2 : make codes meet the requirement of style guide
     7b1411a fix build error
     0c1c6d1 replace old test of SparseAutoencoderFunction with new test
     b3f982c 1 : remove constructor(because it break consistency) 2 : add static functions to initialize weights 3 : fix bug--did not initialize parameters correctly
     b704fb9 fix bug--InitializeWeights is not a member function
     4fbb83b Merge https://github.com/mlpack/mlpack into refactor_sparse_autoencoder
     dd59401 remove boundary check
     af34eb4 adjust format
     0179491 use LogisticFucntion to replace LazyLogisticFunction
     d6f7fa0 use LogisticFucntion to replace LazyLogisticFunction
     12b6d8f 1 : Use LogisticFunction to replace LazyLogisticFunction 2 : fix bugs--Should use HiddenLayer and OutputLayer to defined SparseAutoencoderFunction
     f22e2d5 first commit
     b645986 1 : adjust format 2 : use the layer and activation function from mlpack::nn
     2be611f 1 : const correctness 2 : use the activation function of HiddenLayer to compute the new features
     5c7a6eb use the LogisticFunction from nn
     15a66a1 fix bug--the type should be RectifierFunction
     81169b5 add layer type define, it is needed by FineTuneFunction(pull request 460)
     8a91fc9 add move constructor and move assignment
     09590c3 add test case for move
     487d6b0 refine test case of move
     cb0b5e5 add move constructor and move assignment
     01d5c4b add move constructor and move assignment
     2fd9e2c merge from upstream
     6585b8e 1 : Store performance function 2 : Pass error into Error function 3 : Pass network into Error function
     1dc8262 1 : Store performance function 2 : Pass error into Error function 3 : Pass network into Error function 4 : Add public api to access underlying network 5 : Use perfect forwarding to accept LayerTypes
     69f4514 Error functions now accept error and network parameters
     0a10573 first commit
     308aeac first commit
     72a806f use logistic_function to replace lazy_logistic_function
     fd18cfd omit the old test of sparse_autoencoder
     ed26a46 first commit
     e1bfbc6 1 : can specify more policy 2 : fix bug--type error 3 : fix bug--forgot #endif 4 : change class name 5 : provide alias template--LogisticSparseAutoencoder
     2731bc5 fix bug--const error
     df94d85 1 : fix bug--error range 2 : fix bug--FFN is a type but not data member
     848e20b 1 : change sampleSize to batchSize 2 : able to access batchSize 3 : fix bug--const correctness
     35c5e5b 1 : Use Train function to replace other functions 2 : change sampleSize to batchSize
     eb30b17 Update to latest version.
     ea908de Add functions that allow changing training-time parameters.
     29c0405 Merge pull request #491 from stereomatchingkiss/ann_move
     57b7eb7 Generalize the linear forward method.
     6945c11 Remove extra semicolon which produces a warning when compiling with the -pedantic option.
     7b10efd Don't use a generalized template parameter in the forward function.
     7f4c62b Be sure to serialize and deserialize in the same way!
     095844b Fix closing parenthesis.
     3f79fab Fix bad memory handling and write a test for it.
     805789a Loosen tolerances to increase the probability of successful tests.
     e7dc529 Include correct headers.
     6c42661 Handle an edge case when all the points are the same.
     c505098 Force a minimum value on the diagonal.
     b39417f spelling
     e6d7308 Merge pull request #501 from barak/master
     ee57276 COPYRIGHT.txt, repeated field and lintian tag copyright-refers-to-deprecated-bsd-license-file
     6f8d7a1 Merge pull request #502 from barak/master
     d72958e Fix confusion in error message
     b29997d Merge pull request #504 from andyfangdz/patch-1
     4bbc39a Print correct parameter name.
     941fd20 Widen tolerances so tests succeed more.
     cd15763 Update GetParam<>(parameterName) per marcos
     e6bbde6 Merge pull request #505 from andyfangdz/patch-1
     de6c7b6 Print status while loading.
     0c9829c Fix issue that caused hanging tests: use the intended tolerance.
     a279679 Remove self assignment.
     5e353ff fix bug--forgot to move some values
     e98bc77 first commit
     6e2710f remove non exist and useless headers
     d697ad8 use less generic solution to find out the value type of InputDatatType because travis build cannot pass
     18aad1c use less generic solution to find out the value type of InputDatatType because travis build cannot pass
     61d013a change function name and fix type error
     c1d23bc change template name due to confliction
     1c2af05 change typenmae from FFN to FFNet
     94a3de9 use rvalue reference to pass the value of outputLayer
     12ea51f do not make outputLayer as reference but allow user to access the outputLayer from public api of FFN
     9721cb7 fix bug--lvalue cannot bind to rvalue
     c34e9d3 remove useless variable
     4a9ddb2 remove useless header file
     e481a0a change test suite name
     1a233b4 remove useless preprocessor
     3561ad7 Refactor to allow arbitrary types for Range.
     a7a05d8 Fix typo: we should be checking HAVE_LAPACK not LAPACK_FOUND.
     ab37969 Better error output for ARFF.
     1f6db30 Check number of columns.
     54b2906 A better positive definite constraint strategy.
     33082f0 Merge pull request #451 from stereomatchingkiss/refactor_sparse_autoencoder
     de42a6a Use CMAKE_CURRENT_SOURCE_DIR during files generation.
     4ad39f8 Move the sparse autoencoder into a separate folder and remove legacy code.
     4c5c773 Add the sparse autoencoder to the network traits.
     1896937 Minor style and formatting changes.
     f34ae33 Remove unnecessary element function for the convolutional network and add sparse autoencoder training routine that makes use of the batch size.
     443ecdc Use the sparse autoencoder class for the sparse autoencoder test.
     2f105ec Use the modified sparse autoencoder test file.
     8b4bac7 Remove unused parameter.
     eb2114e Widen tolerances; PositiveDefiniteConstraint seems to change things.
     3b3a8a1 Merge pull request #512 from theranger/master
     47d8a6a Wrap to 80 columns.
     725f7d0 Fix bug where arma_config.hpp wouldn't be installed.
     e7196b5 Update version number.  I forgot to do this!
     5ae0bd9 Loosen tolerance further. This is better-tested than my last attempt and appears to have a very low failure rate.
     b344394 Update history.
     3b5c842 Loosen tolerances to prevent spurious test failures.
     0c62700 Preferentially find ATLAS libraries.
     dea60d5 Update version number for release.
     031604f Add new contributors... how did I forget this?
     fa88d69 Document most recent changes.
     9f1d08a Add new contributor.
     2a57029 Update to latest stable version.
     1a0234f use MSVC14
     ca54300 Merge pull request #518 from theSundayProgrammer/master
     79fd639 Documentation fixes.
     76b2c51 add test case for cube seriliazation
     e4879a7 first commit
     3a6c581 add macro definition to inject serialization of arma cube
     2f86634 Merge pull request #520 from stereomatchingkiss/cube_serialize
     8653bf1 Add test case for empty cubes.
     0603d2e I forget how the XMLs got into the names here, but remove them.
     e4350fb Update build icon location to temporary hack location.
     a1662ef Dash pedantry.
     ec35c2d Use the correct type.
     d86bccc Better documentation.
     2c476c2 Add minibatch SGD implementation.
     e2acdb7 Remove unnecessary includes.
     46e0a23 Add test for MiniBatchSGD.
     9c1bcf6 Switch to C++ headers instead of old C-style.  Fixes #522.
     c751e61 Fix #522, for real this time, by using std::.
     50629c1 Handle MSVC compiler properly; Visual Studio only supports OpenMP 2.5. OpenMP 3.0 is where unsigned indices are supported.
     4e069ab Handle dimensions with no variance better (i.e. ignore them).
     f6dd2f7 Add documentation for change.
     1bdfce1 Minor spelling fix.
     da1207a Add RMSprop implementation.
     b2d87c7 Add test for RMSprop.
     69f97e5 Adjust FFN class so that it works with the mlpack optimizers; Add Train() method regarding the design guidelines.
     f6c27ed Adjust RNN class so that it works with the mlpack optimizers; Add Train() method regarding the design guidelines.
     43f0f13 Adjust CNN class so that it works with the mlpack optimizers; Add Train() method regarding the design guidelines.
     cf3c454 Remove the optimizer and weight initialization rule from each layer.
     a49e0fe Add auxiliary network function to get the size of the network.
     0aeb767 Adjust the network test; Use the simplified layer structure.
     53637f2 Add network auxiliary functions test.
     6205f3e Update CMakeLists file to build the feedforward, recurrent and convolutional neural network.
     8b187fc Remove hard tanh activation function header.
     491123b Add SFINAE pattern for the Weights, Reward and Location function.
     384556e Adjust the activation function test; Use the simplified layer structure.
     efe49b7 Remove debug output.
     a85e4c2 Actually use SGD for the test.
     0341d4d Add better test for mini-batch SGD, and fix bug. Also remove debugging output.
     4f4bd20 Add support for MiniBatchSGD.
     d765779 Set strict = false (use auxiliary memory until the size change) when creating the gradient storage; the default setting changed in armadillo 6.000 from true to false since.
     29dc196 Widen tolerance slightly.
     6e8e873 Allow mini-batch SGD optimizer.
     f3d692c Update HISTORY with MiniBatchSGD changes.
     4ce5e6f Fix -Wmaybe-uninitialized.
     a1486eb Use testData and testResponses to create the test set.
     4a6531b Fix invalid accesses.
     2aedd27 Fix invalid accesses.
     21ef31f add train test split
     e77cb59 refine comments
     e99632f split train test support arma cube
     f7872fb add test of cube
     dc36801 Minor style fix; Add some spaces.
     e671aa8 Add mlpack::backtrace for Ubuntu (C++, Clang).
     c827d59 Commit mlpack::backtrace.
     e9ee6bd Update for mlpack::backtrace
     bce8a34 Delete old backtrace depend.
     aa35528 Add mlpack::backtrace info.
     5f430cd Add mlpack::backtrace header.
     8cbd0d5 Add mlpack::backtrace
     1404c1d Add mlpack::backtrace
     5e2f445 Add mlpack::backtrace
     a215b1b Test the RMSprop optimizer with logistic regression and a feedforward network.
     6257fad Add mlpack::backtrace
     322d384 Merge pull request #1 from mlpack/master
     876de9c Update log.cpp
     95b9320 Dummy constructor
     0a0d68b Dummy constructor
     dc0e553 Include backtrace.hpp only when libs are found
     171c35b Work with mlpack::backtrace only when HAS_BFD_DL..
     7a77a14 Remove unnecessary checks for NULL.
     dd4445d Merge https://github.com/mlpack/mlpack into split_data
     7a7d8e5 change function to class
     8407475 use arma shuffle and linspace to replace old solution
     2fb7ee4 support multiple slice
     30d652d implement serialize
     9176684 implement serialize and remove move constructor and move assignemnt since c++11 should be able to generate them.Remember to specify [ARMA_USE_CXX11]
     2cba975 implement serialize
     7b85147 implement serialize
     f64a84b implement serialize
     9b672da only serialize weights
     2cbd958 implement serialize
     fb7f435 implement serialize
     ad598cc implement serialize
     99a6ada remove move constructor and assignement, let the compiler generated it(Need to define ARMA_USE_CXX11)
     32d4bad remove move constructor and assignement, let the compiler generated it.Remember to define ARMA_USE_CXX11
     2779906 implement serialize
     4f307dd implement serialize
     5782655 implement serialize
     8fd4466 format codes
     debb7ca implement serialization
     07833dd first commit
     dbffa87 fix bug--did not specify range
     0172424 need to specify slice number explicitly
     11e0390 test multi slice and use new api to do the test
     ce37913 fix bug--miss namespce
     c25a2b6 Fix build icon to point to current location of big on ridiculous port number.
     a2e57d6 Merge pull request #527 from stereomatchingkiss/remove_move_implement
     115b38b merge from mlpack
     170c3a2 Merge pull request #2 from mlpack/master
     76e48a5 Update prefixedoutstream_impl.hpp
     e9e7b85 adjust format
     95191cf change namespace to boost and provide overload of serialize
     3517976 remove useless file
     f89396b remove serialization of net
     83f3752 const correctness and adjust format
     55a515b Removed trailing underscores in header guards. Resolved issue #533
     48bfcac Removed Trailing underscores in header guards. Resolves issues #533
     1bedf15 Merge pull request #536 from stereomatchingkiss/ann_serialize
     7af98ad make the type of arma::Row become generic
     bb8507c Serialize all necessary parameter and minor formatting fixes.
     f987048 Merge pull request #3 from mlpack/master
     0616233 fix bug--name confliction
     e75bebb Merge https://github.com/mlpack/mlpack into split_data
     8ffa470 Use CreateNVP from the data namespace.
     6c4eb24 Add -rdynamic if BFD or DL found.
     0bc16ea Change Ubuntu on Linux
     e664f3d Change order in printing backtrace
     f8228a5 Change order in printing backtrace
     0a68185 Added HardTanH Layer and its tests
     78c60a7 Merge pull request #4 from mlpack/master
     bc2bd17 Merge pull request #542 from stereomatchingkiss/fix_name_conflict
     cdb5ac3 Linux specific dependencies
     a0472e4 Platform specific dependencies
     0aad1ee Linux specific dependencies
     7de1645 Changed scope of some functions to private in HardTanH Layer and its tests; some minor style changes
     fc5a4ac Widen tolerance and increase the learning rate.
     cff6a98 Use Nguyen-Widrow method for weight initialization to make the Distracted SequenceRecallTestNetwork more stable.
     f58aa36 Fix compiler is out of heap space error when using MSVC 2015.
     a069a56 Add config to build mlpack using appveyor (windows).
     a3a4656 Fix BatchTrainingTest by splitting into training and test set.
     4fa67e5 Fix namespace ambiguity which causes a build failure on Windows.
     a2deb0c Stabilize DistractedSequenceRecallTest FeedforwardTest test by using a TanHLayer layer.
     f45c17b Add appveyor build status badge and update the build verbosity level.
     ac46f5b Better handling for arbitrary types.
     5f513f0 This isn't really accurate anymore.
     3c4414b Update links for documentation.
     1d7cc4a Refactor to handle arbitrary element types.
     819bef7 Refactor to handle arbitrary element types.
     5006220 Allow BoundType to have multiple template parameters.
     8cfc6f1 Fix incorrect type.
     fcaf28c Add documentation for element type.
     a69871c Refactor trees to allow arbitrary element types.
     38c6d82 Update CMakeLists.txt
     7ae1355 Update CMakeLists.txt
     d5841b1 Fix ridiculous 2 pixel badge size issue. It's an svg, so no quality loss.
     3e366ec Merge pull request #548 from awhitesong/HardTanH
     971213a Remove the single value Forward function to be consistent with the rest of the layer interface and minor formatting fixes.
     fd84f3a Include OpenBLAS when creating the mlpack-windows.zip package.
     256cd1c Fix invalid memory access during XTreeSplit #350
     9467a92 Fix redefinition in RStarTreeSplit::SplitNonLeafNode.
     8921d03 Update XTreeSplit API according to recent RStarTreeSplit API
     1a9c41a Merge pull request #535 from Kirizaki/master
     74593ce Update documentation for executable name changes and soversion bump.
     c75652b Initialize the weights when loading (serialization).
     ea703b9 Added LeakyReLU Layer and its tests; Change the Cmakelists
     50ae8a6 Merge remote-tracking branch 'upstream/master'
     e6f7ffe Merge pull request #559 from awhitesong/LeakyReLULayer
     72e7bff Added XTree to allkfn, allknn, range_search and allkrann
     c3300ce Fix DBL_MAX in r/r*/x_tree_split and r_tree_descent_heuristic
     cf233aa Replace numeric_limits::min() by numeric_limits::lowest() since min() returns the lowest positive for floating point datatypes
     18b10e1 Update allknn, allkran, nrange_search tests.
     3c1142a XTree bugfuxes. Fix a supernode splitting. Fix a supernode creation fault.
     ad97d73 Fix trees order.
     f0e3318 Distinguish between testing and training when optimizing the network model. This is useful for models that act differently when in testing or training mode.
     e3d59d6 Changes to CMakeLists to solve issue #553: Clean build in 32-bit systems
     675101c Added Dhawal Arora as a contributor in copyright.txt and core.hpp
     d9e7edd Merge pull request #565 from awhitesong/copyright
     8a19e56 Merge remote-tracking branch 'upstream/master'
     d91cf4a Changes to solve issue #498: Replacing timer implementation
     930d8e7 Fixed std::logic_error
     5354f5d Incorporated comments from pull requests for issue #498
     f77416e Replace numeric_limits::min() by numeric_limits::lowest() since min() returns the lowest positive for floating point datatypes
     c0886a1 Merge pull request #567 from na1taneja2821/issue-498
     88b9b85 Reimplementing Adadelta with tests
     ff76bc3 Optimizer transition Adam test file
     d673e3e Update the learning rate by calculating the bias correction values.
     b864df8 Build the adam optimizer test.
     386138b Update the default step size (adam optimizer).
     5a8c50d remove support of cube
     8602562 remove support of cube
     eedd99d Fixes bfd.h: config.h must be included before this header
     4551dd0 Refactor for shorter lines.
     eac547a Refactor: remove once-used function, wrap lines.
     086a75d Update names to avoid collisions.
     b3fe523 Merge pull request #575 from u8sand/master
     e93f415 Add comment describing the need for setting PACKAGE and PACKAGE_VERSION.
     91881a9 Add documentation for new backtrace functionality.
     eb1afa1 libxml2 is no longer used; libbfd and libdl are used for the backtrace functionality.
     a3ce99e Remove libxml2-devel from list of Fedora/RHEL dependencies.
     4a2466b Update dependencies for binutils (libbfd and libdl).
     c487c58 Update version in documentation.
     05c6551 I'm tired of this test failing.
     3b5e67f Merge pull request #564 from na1taneja2821/master
     e3d7010 Merge pull request #572 from vasanthkalingeri/stdlogicerror
     ae6689c Use sum of square to calculate theta (see eq. 8: sum(x^2))
     b72c0b8 Merge https://github.com/mlpack/mlpack into split_data
     80943dd Add some paranoia: what if sizeof(void*) isn't 4?
     e5076db Fix bug #358 (CosineTreeTest/CosineNodeCosineSplit test sometimes fails, only on i386)
     121bbaf This commit allows to avoid splitting a node which owns a dataset and has children with the same dataset.
     1ed8c18 Use RTreeDescentHeuristic. Add documentation.
     3fbabd7 Minor style fixes.
     322ff1c Build the AdaDelta test.
     5641b4d Remove unused result variable from the AdaDelta test.
     74d499a Remove unused ann optimizer; For more information take a look at #555.
     a4e3260 Remove reference to old ann optimizer header.
     ca25190 Fix spacing.
     3f3d018 Converts CMake commands to lower case
     24a224a Merge pull request #584 from ajjl/CMakeLowercaseCommands
     5ca7299 Handle loading and saving HDF5 datasets correctly.
     2330419 Removes CMake block termination arguments
     1f75ba1 Manually set flag for _WIN64
     14273c0 Merge pull request #586 from ajjl/CMakeBlockendArguments
     758d277 Merge pull request #589 from cloudhan/warningC4267-patch
     3253027 lsh_test tests if recall increases with L
     501ffa5 The arma::hdf5_binary type already transposes on load and save.
     7199297 Fix misspelling.  Oops!
     3150b12 Fix error in adjustedScore calculation.
     63a7f62 Adding complete implementation of dropconnect layer with all the suggested changes.
     6eba278 small style changes
     5775f3f small style changes in the test case file
     f13428a Minor style fixes.
     11b4b5e Use template parameters for the empty layer class (placeholder) to be usefull for all layer and minor style fixes.
     e3b4189 Merge pull request #590 from MarcosPividori/master
     24e6092 Merge branch 'master' of https://github.com/lozhnikov/mlpack
     7d1bcd8 Adding name to the list of contributors
     5af7dd9 Merge pull request #591 from palashahuja/master
     998cd7f Loosen tolerance a bit so this test randomly fails less often.
     3988fc6 Bug fix for Elkan?
     04b0c08 Minimal fix for bug in Elkan implementation.
     c14e81a Fix link.
     a0f1dd5 Fix dual-tree k-means runtime bug.
     9e4e126 Merge pull request #592 from kno10/patch-1
     0031a82 Adds two tests for LSHSearch
     1e1bc6f Use BOOST_REQUIRE_* instead of BOOST_CHECK_*
     799cafe Add testing for LSHSearch and refactors code
     68e2cbf Ensure the last batch size isn't zero, to avoid division by zero before updating, when using a batchsize that fulfilled the constraint: (numFunctions % batchSize) == 1.
     6359987 Add simple test that to ensure the batch size isn't zero.
     66cb285 Divide the root square of the second decay value by the first decay value; clarification of the beta parameter.
     fb75f38 Merge pull request #581 from lozhnikov/cosinetree
     6a70cfc Minor style fixes.
     e14c246 Document cosine tree fix.
     9f72e89 Fix unnecessarily long line.
     1708dfd Add 3 more recall-based tests for LSHSearch
     bced046 Add failsafe to numProj test of LSHSearch
     289782e Style Fixes
     93a42c8 Merge branch 'master' of https://github.com/mlpack/mlpack into issue261/LSH_test
     b0e5d74 Fix failsafe for numTables recall test
     18c925a Style changes
     9561fe2 Add Yannis Mentekidis to contributors
     5bc514c As pointed out in the program info; the reference set will be used as both the reference and query set if no separate query set is specified.
     c838566 Style Fixes In lsh_test
     f229e4e Separate tests into Boost test cases
     9ef75a4 Whitespace fixes
     f0675d7 Merge pull request #605 from mentekid/issue261/LSH_test
     638381a Merge remote-tracking branch 'upstream/master'
     65df1ea Fix trees ordering.
     bc6b9e7 Style fixes for LSH test.
     ba826b1 Make cleaner Armadillo expression.
     d0708a5 Add function to get the input size of a given network.
     f941837 Add test case for the layer input size and network input size function.
     b33964b Add auxiliary function to set the network weights using a given initialize rule.
     07289ba Add test case for the layer and network weights function that uses a given initialization rule to initialize the weights.
     f4b3464 Add confidence parameter to enable custom class transformation.
     0e8d776 Make sure, the gradient isn't empty, before propagating the gradient parameter through the network.
     592fdcd Reorder the parameter of the DropConnectLayer to avoid reorder warning.
     fb3994c Add option to set the input parameter when calculating the gradient.
     c0dbd32 Add ConstantLayer class that outputs a constant value given any input.
     460e326 Add LogSoftmaxLayer class; the log softmax loss layer computes the multinomial logistic loss of the softmax of its inputs.
     7e1e1de little change
     3f2d6ab name added
     d1c2ae0 branch conflict resolved
     78f6ac2 Merge pull request #611 from ranjan1990/SGD
     63be94e use constexpr to replace static const
     eeba6bd Merge pull request #612 from stereomatchingkiss/refine_log_softmax
     a955ee1 Clarify section name.
     6ccf865 Add a tutorial for the formats.
     243ace7 Document the DatasetInfo class more comprehensively.
     cb3c489 Don't set the random seed!  That is done at higher levels.
     a38608b Add SampleInitialization and different type of initialization.
     b8bb079 Refactor RefinedStart to give centroids not assignments.
     c2e0107 Switch default initialization strategy.
     01cf94c Test SampleInitialization.
     8d77f42 Update tutorial for k-means. This considers the new API for initialization strategies.
     59d07ef Minor formatting fixes (remove tabs).
     78f53da Print backtrace after newline in fatal message.
     d6d38a1 Fix HMM initial state estimation.
     11d6539 Update history.
     b08ff03 only initialize variable once
     1417d45 Add ReinforceNormalLayer class which implements the REINFORCE algorithm for the normal distribution.
     0fe4172 remove useless data and function
     595ff70 1 : remove seed variable 2 : do not store testRatio, pass in by function
     75e946e change class to function
     7723421 refine test case
     b5790a0 refine comments
     e0e7ba5 Handle case when dataSeq.size() == 1. Thanks warpuv for pointing this out!
     77532c8 Merge pull request #614 from stereomatchingkiss/refine_log_softmax
     f3db427 MinGW: use lowercase windows.h
     f73d898 Merge pull request #615 from xantares/patch-1
     79bf330 Update documentation.
     15b5e65 Add recent tutorials.
     8ca7cf2 Fix doxygen issue.
     1140d93 Fix missing \endcode.
     08ffa1b Fix name of tutorial.
     c9c1712 Merge branch 'master' of https://github.com/Falit/mlpack into Falit-master
     e0e9d04 Update version macros to retain underscore versions.
     3781ccb Fix unusual bug that pruned too tightly.
     637809f Merge branch 'master' of https://github.com/mlpack/mlpack
     24f7626 Update HISTORY to note changed version identifiers.
     b8f53a4 Update documentation.
     62a9489 Compile in debugging mode, in the hopes that things will be faster.
     e047bcf We have to tell msbuild to compile in debugging mode also use the custom appveyor logger to push MSBuild warning and errors in real-time to build console. Let's see if that makes things faster.
     6de06f0 Add msbuild /bigobj flag to the appveyor config to avoid 'object file format limit' error.
     2e33c5c Switching back to release mode; also remove the bigobj flag since it should be added to the CMake file.
     4fa02a0 Run mlapack_test through ctest
     06c34f7 Add an option to disable tests
     0c945f3 Manually calculate starting state probability.
     f13409c Rename mlpack_allknn/mlpack_allkfn to mlpack_knn/mlpack_kfn
     6982488 Merge pull request #616 from xantares/testing
     31dd6c0 Document new CMake options.
     2f9ad40 Make mlpack_allknn by copy
     5702b9f Copy mlpack_knn/mlpack_kfn after they have been built
     6f213a3 add ARMA_USE_CXX11 if the compiler is msvc
     c92dd2f remove suggestions of add ARMA_USE_CXX11
     6f9a29c use WINDOWS_EXPORT_ALL_SYMBOLS
     9cd5bec Add an option to disable CLI executables
     c6eb4d7 Fix a cast in dt_utils.cpp
     c5573b2 Merge pull request #621 from xantares/fix-dt_utils-cast
     0e6f351 Merge pull request #618 from darcyliu/rename
     3d65139 Update HISTORY to note mlpack_allknn and mlpack_allkfn change.
     cf2b625 Remove debugging output.
     3991cc2 Merge branch 'cli_option' of https://github.com/xantares/mlpack into xantares-cli_option
     d6c4a52 Remove extra newlines.
     2564cf4 Update documentation.
     28ae007 Merge branch 'xantares-cli_option'
     c5aa370 Fix configuration when BUILD_CLI_EXECUTABLES=OFF
     ee3c27a Merge pull request #625 from xantares/fix_cli_opt
     8a249cf Rename main program files.
     a6f20af Update comment for accuracy; I should have been more specific when I wrote it.
     7e0e6ea My editor auto-removed trailing spaces.
     56b53a0 Remove unnecessary comments from tests.
     1d7ea16 Deprecate old names; add new names.
     6be725a Update documentation for changed names.
     efb64f2 Removes trailing whitespaces at end of lines
     0e583d2 Remove non-existent man page targets
     64c049e Rename AllkFN to KFN, and AllkNN to KNN.
     c334b99 Merge pull request #627 from darcyliu/bugfix
     d2e3534 Merge pull request #626 from ajjl/removeTrailingWhitespace
     54c7f22 1 : fix style 2 : remove useless header 3 : reduce temporary variable
     b7a01eb BUG: Misplaced parenthesis in comparison of absolute values
     ebf77f8 Merge pull request #628 from ajjl/parenthesisBugFix
     03a34ed Replace static SplitType in the RectangleTree by an instantiated object.
     b08ae02 Merge pull request #523 from stereomatchingkiss/split_data
     c20195c Move <tuple> to standard include.
     410bed0 Move SplitData to data namespace.
     35149da Style fixes.
     213a04d Add another test, and check that we aren't duplicating points.
     6008c15 Use the RectangleTree::Split() method instead of the friend class. Move the normalNodeMaxNumChildren to the XTreeSplit.
     f0f9fff Merge remote-tracking branch 'upstream/master' Fix conflicts
     aa7e513 Merge pull request #622 from xantares/windows_shared_libs
     a59d125 Better documentation of BUILD_SHARED_LIBS option.
     7264d37 Add a bit of documentation for what MLPACK_EXPORT is and why it is there.
     d47933c Fix mlpack_export.h include
     4c8a8d1 Merge pull request #634 from xantares/fix-mlpack_export-include
     52423b8 Remove a pointer to the node from the SplitType class.
     4208f38 Move splitHistory to XTreeSplit. Update serialization of the RectangleTree class.
     019680f Combine SplitType's constructors.
     93eb34c Merge pull request #556 from lozhnikov/master
     d502539 Move mlpack_export.h to mlpack_export.hpp.
     27ebca1 Fix a mistake in metric policy.
     1f562a1 Merge pull request #637 from MarcosPividori/master
     4536910 Force positive-definiteness when training a Gaussian.
     2744628 Better handling of NaNs.
     3e4f3ca Add --random_initialization for mlpack_hmm_train.
     9866203 Don't forget the period in the output.
     2114358 Add MultiplyConstantLayer which multiplies the input by a non-learnable constant.
     cc7bac3 Add NegativeLogLikelihoodLayer class which is useful to train a classication problem with n classes.
     b69c6dc Add the VRClassRewardLayer class which implements the REINFORCE algoritm for classification models. To be precise, this is is a Variance Reduces classification reinforcement learning rule.
     be43684 Add GlimpseLayer class which takes an input image and a location to extract a retina-like representation of the input image.
     fe69d33 Remove debug message.
     4d8347a Add Recurrent Model of Visual Attention (RMVA) implementation.
     d1650e4  Removes trailing whitespaces at the end of lines.
     989dd35 Include split_data.hpp file into the build process, so that it is available afterwards.
     f1bf339 Remove unused output parameter.
     39eefde Add RMVA class and function documentation.
     dd136db Properly use Enum type.
     9b42c22 Merge pull request #641 from MarcosPividori/master
     9b811f9 Typo fix in knn_main.cpp
     36b7316 Merge pull request #643 from dasayan05/master
     c82c747 Properly use Enum type.
     eef40b9  Remove duplicated code for traversal info.
     b4ee954 Remove unnecessary include.
     f55427d Merge pull request #645 from MarcosPividori/master
     ade7fa0 Merge pull request #646 from MarcosPividori/traversal-info
     8acc4ce Remove trailing underscores in header guards as discussed in #533.
     cd7f063 Add CMakeLists file to build the RMVA.
     c15541b Remove CMakeLists file; do not build the RMVA code.
     04fe0d5 Use n_rows and n_cols to define the matrix size instead of arma::size().
     4fae385 Deprecated arma function replaced by new arma constant
     5d1723d Merge pull request #648 from dasayan05/PR1
     6f6173c Update documentation for changed names.
     a35c390 add cli executable for data_split
     c68c3b9 vc2_test.csv wasn't in proper csv format
     fe1b6b9 fix performance issue--forgot to move name parameters
     de4226c change parameter flags and names for consistency with other programs #discussion_r64743312
     290977d Merge pull request #652 from dasayan05/test
     2cad593 add TrainTestSplit without label
     c8a60b2 rename TrainTestSplit to Split
     ec1c46a Merge pull request #653 from stereomatchingkiss/split_data_performance_issue
     e41b3db add data split executable without labels + tests
     706f0ee Fix error, a balltree should use ballbounds instead of hrectbounds.
     9e5d9c2 Set proper template order. To work with binary space trees, the metric template parameter must be the first.
     c0cdb8b Proper template order to BallBound.
     903be7e Fix error. kdtree where should be a balltree.
     8285c31 Change template name to be similar to the definition of hrectbound.
     ef01c05 fix directive name for split_data.hpp
     3d1ed0f Use add_cli_executable to control CLI executables build; more information take a look at #619.
     2fe9e82 Merge pull request #655 from MarcosPividori/fix-ball-tree
     1dad2b6 Fix spacing.
     9e88669 refine style and detail test
     0e61fe2 1 : fixing issue #658, work with non-transpose data, do not work with transpose data yet
     29f0536 remove useless codes
     9aac1dc use algorithm to replace for loop
     8ad3b90 Merge pull request #1 from mlpack/master
     e36eec5 Document the state of loading sparse matrices.
     a340e23 fix bug--isdigit will not treat 1.004e10 as digit
     56a0361 support transpose mapping
     0634d4a fix format--place const before variable
     5d964a1 fix bugs--may read invalid stream
     6eb1723 remove debug message
     5eaa2a9 reduce copy
     4ea13b0 add new test cases
     1beb58d A first simple test.
     3ee8906 fix bug--forgot to put the functions into nameless namespace
     02e31b3 Better Doxygen formatting.
     3e75b48 make TransPoseTokens as rcurtin suggested, should not use nameless namespace in header file
     6877607 Implement B_aux according to what was discussed in #642.
     ba4c5bb Remove unused information from neighbor search stats.
     79d07f2 Remove unnecessary include
     b3fc468 Adapt Keon's test into something harder.
     e94896d Adds code that gives controllable access to LSH projection tables
     f3c4939 Adds code that gives controllable access to LSH projection tables
     934fe08 Adds code that gives controllable access to LSH projection tables
     e6d2ca7 I'm not sure what line width Pari used, but it wasn't 80 columns.
     4773efb Fixes naming conventions of accessors
     b067e89 Changes LSHSearch.projections from std::vector<arma::mat> to arma::cube
     7ebf1da Changes LSHSearch::projections from std::vector<arma::mat> to arma::cube
     ff7877d Add a templated version for BOOST_CLASS_VERSION().
     6958cc0 We actually need to wrap mlpack::data::SecondShim<> objects.
     1474050 Include new serialization version macro.
     81b6b4f Use n_slices not size() to fix correctness.
     f989f1f Refactor Serialize(), add backwards compatibility, and update tests.
     5546ebc Add Classify() functions and tests.
     dca52fd Add --output_probabilities_file option.
     eba4f99 Fix spacing and make method const.
     2417a19 add more program info of split
     ac016fe fix tests for split data
     06fdfa8 Adds new constructor to LSHSearch and merges Train() and BuildHash() methods
     e0b6ce7 Merge branch 'LSHTableAccess' of https://github.com/mentekid/mlpack into mentekid-LSHTableAccess
     b30e697 Remove BuildHash(); simpler cube default.
     d3e3c54 Minor code cleanups and style fixes.
     8ed22ce We can't remove Projection() because we'd break our versioning principles. So we'll have to wait until mlpack 2.0.1 to remove it... :(
     e3a23c2 Update HISTORY.
     181792d Merge pull request #650 from keonkim/master
     6a94eb6 Update some documentation, change --input_labels to --input_labels_file.
     281555a Merge pull request #668 from MarcosPividori/knn-bounds
     dde781f Marcus thinks this will fix the Windows build... let's find out.
     2629a4f Merge pull request #670 from rcurtin/master
     ee95e20 Merge pull request #660 from stereomatchingkiss/fix_mapping_issue
     a2440f1 Merge pull request #661 from rcurtin/keontests
     940a2b5 Refactor for faster assembly of secondHashTable.
     06cae13 Update aged Fedora documentation.
     5b8fdce Revert "Refactor for faster assembly of secondHashTable."
     281fb8a Merge branch 'master' into mlpack-2.x