gsps Activity
A Java library implementing Klir's General Systems Problem Solver
Brought to you by:
smiz
Activity for gsps
-
smiz committed [r184]
More work on classification
-
MOre work on classifier
-
Ugh
-
More work on classifier
-
More work on classifier
-
More work on classifier
-
Added character based classifier
-
Fixed the overfit problem?
-
Trying to correct for overfitting
-
More work on sequence classifier
-
More work on sequence classifier
-
New sequence classifier
-
More objective function fun...
-
This objective function stuff is tricky
-
More work on objective function
-
More work on objective function
-
More work on objective function
-
More changes to objective function
-
Updates to seq classifier
-
Update to objective function and test code
-
More changes to the objective function
-
New objective function
-
More tinkering with objective function
-
Another update to sequence classifier
-
Yet another model builder
-
Yet another objective function
-
Removed weighting from fitness function
-
More tinkering with sequence classifier's objective function
-
New scoring function for sequence classification
-
Another sequence classifier
-
Performance improvements for addTimeSeries and seq_probabilies in the mask class
-
Speedup of addTimeSeries
-
Looking for a good objective function
-
Code cleanup in classifiers
-
Forgot to add a source file
-
New sequence classification method
-
Update to classification example
-
Cleanup of classifier example
-
Fixed problem with reject rate calculation
-
Change in token selection
-
Update to seq classification example to use new token selector
-
More work on sequence classifier
-
Update to text classification method
-
Ugh
-
Fixed missing input error in Automaton
-
Small performance improvements (?) and updates to examples
-
Added ability to save models
-
Work on seq classification example
-
Performance improvements to speedup automaton parameter search
-
Added hill climbing procedure for optimizing automaton
-
Performance improvements for calculating sequence probabilities given using an automaton
-
More work on evolutionary algorithm
-
Update to evolutionary search algorithm
-
Added capability to add a hidden state to a mask
-
Update to split function to keep obervation counts
-
Code cleanup and added save at improvements to the genetic algorithm
-
Code cleanup
-
Working(?) text classification system added to the examples
-
Added text classification example
-
Added functionaliy for calculating sequence log likelihoods
-
Updates for more easily building binary classifiers
-
Put negative examples back into classifier
-
Report Q for false positive
-
Forgot trainer
-
Removed defunct java files from hmm
-
Reintroduced negative examples. Thes appear to help with some text classification tasks
-
Added parallel for loop to text classification test
-
Removed ill advised negative training
-
More numerically robust test for relevance
-
Eliminated estimate of P(s) from P(C|s) calculation
-
Text classification using string likelihood with uncertainty measure
-
Classification using likelihood of strings
-
New tree building strategy looks at prevelance of substrings
-
Stop classifier scoring at first point of certainty in a string
-
Test accepts test or train argument
-
Changed score to average error per string
-
Mistake in test.cpp to pick wrong file for testing
-
Changed word file name
-
New error function that favors reduction of larger contributors first
-
Training base on a more 'rational' objective function
-
Fixed problem in test code that cause report of zero FP'
-
Added example application that classifies reuters articles
-
Improved training for classifiers
-
Remove MPI C++ bindings from test cases
-
Remove C++ MPI bindings
-
Added traiiner for classification
-
Fixed bug in MPI code
-
Alterate writes to file so premature end doesn't corrupt solution
-
Added mpi support
-
Forget a file
-
Improved(?) evolution
-
Performance improvements
-
Changed mutation rule
-
Pulled out log likelihood function
-
Option to create filename and some convergence improvements
-
Code cleanup
-
Added error checking for score function
-
More performance improvements
-
More work on speedup
-
Continuing with performance improvements
1 >