Fast Artificial Neural Network Library


To get the latest stable release, go to the Download page, for earlier releases go the SourceForge downloads and for unreleased versions go to the git repository.

FANN 2.2.0 – 2012-01-24

  • Added Sarprop training
  • Added fann_create_train for creating an empty training data struct
  • Added fann_copy for copying an ANN
  • Added cascade_min_out_epochs and cascade_min_cand_epochs to improve cascade training
  • Added extra checks when training, to ensure that data and network input and output sizes matches
  • Added Visual Studio 2010 solution
  • Added support for 64bit architecture
  • Cleanup in sources
  • Moved source from CVS to GIT

FANN 2.1.0 – 2007-02-01

  • Added fann_create_train_from_callback for creating training data from a callback function
  • Amended fann_create_train_from_callback for dynamic link library compatibility
  • Added fann_get_activation_function and fann_get_activation_steepness matching the set functions
  • Added a set of functions to allow restricted access to neural network layout, connections, and weights.
  • Changed the create_array functions to indicate that the layers array parameter is const i.e. input only
  • Added FANN_API calling convention to fann_get_bit_fail and two cascade functions for dll compatibility
  • Added FANN_EXTERNAL/FANN_API calling convention to previous extension functions for dll compatibility
  • Removed warning from fann_print_connections with explicit casts
  • Added fann_set/get_user_data to allow user programs to refer to their own data during callbacks
  • Added documented C++ wrapper and sample C++ program
  • Fixed bias count for shortcut networks in fann_get_layer_array and fann_get_bias_array
  • Minor fixes of FANN_EXTERNAL/FANN_API for windows dll compatibility
  • More advanced scaling functionality
  • Added advanced scaling functionality to fann_cpp.h
  • Additional documentation and fix of FANN_EXTERNAL for create train callback in fann_cpp.h
  • Fixed problems with FANN_GAUSSIAN and FANN_GAUSSIAN_SYMMETRIC activation functions
  • Added semi Widrow + Nguyen to initialization of cascade weights
  • Fixed problem with RPROP, which makes it go suboptimal in the beginning
  • Changed definition of MSE, so that it calculates a value pr output, and not a sum for all outputs. (MSE = MSE/ann->num_output)
  • Added more benchmark data
  • Added possebility to make benchmark as average of several
  • Changed so that RPROP and Quickprop may not alter a weight to more than 1500

FANN 2.0.0 – 2006-01-06

  • Added Delphi bindings
  • Cascade training
  • Momentum parameter
  • Steepness and activation function available for single neurons
  • New documentation
  • Removed some depricated functions
  • New set of fann_create functions
  • Altered the way the callback function works
  • Training can be stopped by counting the number of datasets that fail instead of just MSE
  • Fixed a memory leak when copying training data
  • More activation functions
  • Enums used instead of integers for several parameters.
  • Possebility to set default logging facility
  • All parameters are saved to configuration file and not just the parameters needed to execute the ANN
  • More readable configuration file
  • Visual Studio .Net Project Files
  • Borland C++ Builder 6.0 Project files

FANN 1.2.0 – 2004-10-09

  • Fixes for better compability with different compilers
  • Shortcut connections that skip layers
  • Changed version of save fileformat
  • fann_print_connections function
  • More activation functions
  • Batch training
  • RPROP training (default)
  • Quickprop training
  • fann_get_training_algorithm and fann_set_training_algorithm functions
  • fann_test_data function
  • fann_print_parameters function
  • Renamed some functions and variables to avoid confusion of terms (kept deprecated functions)

FANN 1.1.0 – 2004-03-31

  • Error checking when allocating memory
  • Debian package
  • RPM package
  • Added pkg-config
  • Deallocate correctly in case of an error
  • Renamed fann_get_error to fann_get_MSE (keeping old one for now)
  • Renamed fann_reset_error to fann_reset_MSE (keeping old one for now)
  • Rearanged some code for easy access by the newly created fann php module
  • Defined more activation functions and documented the old (in fann_activation.h)
  • Customize error output
  • Added fann_set_error_log, fann_shuffle_train_data, fann_merge_train_data, and fann_duplicate_train_data, fann_init_weights
  • Added python bindings
  • Created DocBook XML documentation, including API reference
  • Rearanged files for easier development
  • Added MSVC++ project files

FANN 1.0.5 – 2004-01-06

  • Added a new constructor
  • Added a callback function for output during training
  • Changed compile parameters for further performance
  • Error value not set to zero after training
  • Thanks to Evan Nemerson for some of these patches

FANN 1.0.4 – 2003-12-11

  • Bugfix concerning default activation function

FANN 1.0.3 – 2003-12-09

  • Added FANN_SIGMOID_STEPWISE activation function (faster)
  • Moved files around and altered some files for easyer benchmarking
  • Added a benchmark pdf file, benchmarking the new behavior.
  • Better compability with non gnu compilers (should work off the shelf)

FANN 1.0.2 – 2003-11-13

  • Added configure script
  • Moved files around, to make the transition to configure more easy
  • Removed other versions of documentation than pdf

FANN 1.0.1 – 2003-11-03

  • Bugfix concerning steepness parameter
  • Moved some files around
  • Added the changelog

FANN 1.0.0 – 2003-11-01

  • Initial release of full functioning library

No Comments Yet

Leave a Reply