| Version | Current Version Number |
|---|---|
| Release | 2.40 |
| Beta | 2.42.0 |
Aqwam's Machine, Deep And Reinforcement Learning Library (Aqwam-MDRLL)
Author: Aqwam Harish Aiman
Email: aqwam.harish.aiman@gmail.com
LinkedIn: https://www.linkedin.com/in/aqwam-harish-aiman/
YouTube: https://www.youtube.com/channel/UCUrwoxv5dufEmbGsxyEUPZw
View the documentation here: https://aqwamcreates.github.io/DataPredict/
By using or possessing any copies of this library or its assets (including the icons), you agree to our Terms And Conditions.
For information regarding potential license violations and eligibility for a bounty reward, please refer to the Terms And Conditions Violation Bounty Reward Information.
DataPredict - Development Pause
After over three years of continuous development, I’m placing DataPredict into a maintenance / pause state.
The library is stable, complete, and usable in its current form, and all released features will remain available as-is. This is because I’m currently focusing on research and other personal projects, and I have determined that the current DataPredict's features are more than enough to cover most game-related use cases.
DataPredict's development may resume in the future when there’s a clear technical or practical reason to do so.
Thanks to everyone who used, tested, or followed the project since its first release in 2023.
Number of algorithms per model type:
| Model Type | Purpose | Count |
|---|---|---|
| Regression | Continuous Value Prediction | 17 |
| Classification | Feature-Class Prediction | 13 |
| Clustering | Feature Grouping | 10 |
| Deep Reinforcement Learning | State-Action Optimization Using Neural Networks | 26 |
| Tabular Reinforcement Learning | State-Action Optimization Using Tables | 17 |
| Sequence Modelling | Next State Prediction And Generation | 3 |
| Filtering | Next State Tracking / Estimation | 4 |
| Outlier Detection | Outlier Score Generation | 4 |
| Recommendation | User-Item Pairing | 5 |
| Generative | Feature To Novel Value | 4 |
| Feature-Class Containers | Feature-Class Look Up | 1 |
| Total | 104 |
-
For strong deep learning applications, have a look at DataPredict™ Neural (object-oriented, static graph) and DataPredict™ Axon (function-oriented, dynamic graph) instead. DataPredict™ is only suitable for general purpose machine, deep and reinforcement learning.
-
Uses reverse-mode automatic differentiation and lazy differentiation evaluation.
-
Includes convolutional, pooling, embedding, dropout and activation layers.
-
Contains most of the deep reinforcement learning and generative algorithms listed here.
-
-
Currently, DataPredict™ has ~93% (97 out of 104) models with online learning capabilities. By default, most models would perform offline / batch training on the first train before switching to online / incremental / sequential after the first train.
-
No dimensionality reduction algorithms due to not being suitable for game-related use cases. They tend to be computationally expensive and are only useful when a full dataset is collected. This can be offset by choosing proper features and remove the unnecessary ones.
-
No tree models (like decision trees) for now due to these models requiring the full dataset and tend to be computationally expensive. In addition, most of these tree models do not have online learning capabilities.