DataPredict™ [3 Years / Release 2.40] - Revenue & Game Optimization Using Machine Learning, Deep Learning And Reinforcement Learning (100+ Models)

Resources Community Resources
, , , ,
1127

Jeez. What’s with very high number of views?

image
image393×287 16.8 KB

1 Like
1128

Because your library is considered best.
Even AI Overview in google search label your library as “General Purpose/Variety”

1129

Wait, did you came from the search engine results or by looking through the DevForum’s search tool?

1130

No.
But the search engine says your library is considered the best.
Search “Roblox best ML library only 1.”, based on what AI over review gives me

1131

THANK U SM TYSM JUST DISCOVERED THIS

1132

Your welcome, I guess.

Also, just to remind you all, this library is currently under maintenance mode until further notice since the algorithms in here covers mostly all the use cases.

So, no new features and future updates will likely involve bug fixes.

2 Likes
1133

will the api ever change?

1134

Extremely unlikely. I have perfected it for over three years.

You can expect these constructors’ and functions’ behaviours would never change:

Constructors

  • .new()

Functions

  • :train()

  • :predict()

  • :getModelParameters()

  • :setModelParameters()

1 Like
1135

Currently, I have added a few more models to the current beta version and I will add a few more into the same current beta version. However, unlike the previous beta versions, the future beta versions would likely be released more slowly and does not instantly get pushed to the release version for “stability”. So, you may see slight discrepancies between the DataPredict’s Model API reference and the current DataPredict’s model list.

I won’t post an official update announcement for now for the beta version.

Thanks for staying around!

1136

Beta Version 2.38.0

Added

  • Added these models under the “Models” section:

    • OrdinaryLeastSquaresRegression

    • RecursiveLeastSquaresRegression

    • WeightedLeastSquaresRegression

    • IterativeReweightedLeastSquaresRegression

Changes

  • Renamed NormalEquationLinearRegression to RidgeRegression under the “Models” section.
1137

To repeat again, this beta version does not include with its own release version. I plan to take things more slower for better “update” trade-off from the previous release version and the next release version.

1138

Beta Version 2.39.0

Added

  • Added these models under the “Models” section:

    • GaussNewtonRegression

    • LevenbergMarquardtRegression

    • LinearRegressionCovariancePreconditionedVariant

    • SupportVectorRegressionCovariancePreconditionedVariant

    • SupportVectorMachineCovariancePreconditionedVariant

1139

Release Version 2.38 / Beta Version 2.40.0

Solvers

  • Added these solvers:

    • Gradient

    • ConjugateGradient

    • GaussNewton

    • LevenbergMarquardt

    • IterativeReweighted

    • GreedyCoordinate

    • RandomCoordinate

Models

  • Added “Solver” parameter to the new() constructor for these models:

    • LinearRegression

    • SupportVectorRegression

    • SupportVectorRegressionGradientVariant

    • PoissonRegression

    • NegativeBinomialRegression

    • GammaRegression

    • QuantileRegression

    • BinaryRegression

    • SupportVectorMachine

    • SupportVectorMachineGradientVariant

    • OneClassSupportVectorMachine

    • ConditionalRandomField

  • Added these models:

    • OrdinaryLeastSquaresRegression

    • RecursiveLeastSquaresRegression

    • WeightedLeastSquaresRegression

1140

Apparently, I just benchmarked my new gradient solver object from the latest release version against the usual non-solver gradient calculations from the older release version…

Apparently, I have made the library 20% more faster for the gradient calculations thanks to the solver’s cache capability, and I expect the performance for the other mathematically-heavy solvers to be huge after the first training iteration.

1 Like
1141

Beta Version 2.41.0

Added

  • Added these solvers under the “Solvers” section:

    • GaussSeidel

    • Jacobi

  • Added these models under the “Models” section:

    • PartialLeastSquaresRegression
1 Like
1142

Wait… Why can’t I change the title?! Anyone knows what’s going on here?

1143

Release Version 2.40 / Beta Version 2.42.0

Added

  • Added the ability to stop the iterations early if the cost value is a nan (not a number) value for all models that uses IterativeMethodBaseModel under the “Models” section.

  • Added autoResetConvergenceCheck parameter to the IterativeMethodBaseModel’s new() constructor under the “Models” section.

1144

High-Value Project Tutorials Update

Previously, I had used LinearRegression models for any systems involving time-to-leave prediction, completely forgetting the fact that this model would produce negative values that would be considered invalid for time-based values. As such, I have replaced it with GammaRegression for better representation and naturally handles right-skewed distribution.

I recommend that anyone who use LinearRegression for this application to replace it with GammaRegression for better accuracy, stability and representation.

1145

I have chosen a more controversial title from:

DataPredict™ [3 Years / Release 2.40] - Revenue & Game Optimizations Using Machine Learning, Deep Learning and Reinforcement Learning (100+ Models)

to:

DataPredict™ [3 Years / Release 2.40] - Don’t Let Your Game Die By Using Machine Learning, Deep Learning And Reinforcement Learning (100+ Models)

Let the war begin!

1146

Hi! I havent used this yet, but itll be fun to learn!

Thanks for making this!