DataPredict
| Developer(s) | Aqwam Harish Aiman |
|---|---|
| Initial release | March 24, 2023 |
| Stable release | Release Version 1.14
/ March 24, 2024 |
| Preview release | Beta 1.29.0
/ March 24, 2024 |
| Repository | github |
| Written in | Lua |
| Engine | |
| Type | Machine learning library |
| Website | aqwamcreates |
Search DataPredict on Amazon.
DataPredict[1] (formerly known as Aqwam's Machine And Deep Learning Library) is a machine and deep learning library specifically designed for Roblox. It offers a range of features and includes more than twelve models. The library aims to provide a user-friendly API, although it may sacrifice some customizability. Additionally, it offers API design standards for users who wish to create their own models and optimizers.
History
AR-MachineLL
The prototype version of the library, called Aqwam's Machine Learning Library (A-MachineLL)[2], was initially developed in January 2023. It gained attention when shared on Roblox's DevForum, receiving positive feedback before being deleted. The creator cited the library's outdated nature, along with API design and calculation issues, as reasons for its removal.
DataPredict
DataPredict was introduced the next month as a replacement for the prototype version. It featured notable improvements, primarily in its object-oriented design. Initially, it faced criticism due to a lack of documentation, but the creator apologized to the community and made subsequent improvements. As of now, the library has garnered over 15k views and 145 likes on Roblox's DevForum.
Features
User-Friendly API
DataPredict provides pre-built models and optimizers, eliminating the need for users to create models from scratch. This approach simplifies the training process.
Retrainable Models
Users can retrain the models as many times as desired, allowing for adaptability based on changing requirements.
Model Parameters Saving and Loading
The library enables users to extract and store model parameters in data storages such as Roblox's DataStore service. These parameters can be loaded back into the model when the original settings are restored. Different models have different model parameters, which is addressed in the library's documentation.[1]
Variety of Models
DataPredict offers a diverse range of models, including (reinforcing) neural networks, support vector machines, and linear regression. The models also cover a number of functions and parameters, such as adding an optimizer.
Supports Deep Reinforcement Learning Algorithms
DataPredict contains a number of popular deep reinforcement learning algorithms, including Proximal Policy Optimization and Deep Q Learning. It also offers experience replays for temporal-difference algorithms.
Limitations
Platform Limitations
In one of the posts on Roblox's DevForum, the community cited the hardware limitations for the lack of machine and deep learning adoption in games. Nevertheless, they expected that these kinds of libraries will be adopted more in the future.[3]
Sequential Model Performance
Models like LSTM and RNN have longer training times due to their sequential nature. Due to the large number of calculations needed, it may freeze or crash Roblox Studio when training them.
Dependencies
MatrixL
MatrixL (formerly known as Aqwam's Matrix Library) is a matrix library specifically designed for Roblox. It was developed to avoid potential legal issues associated with other existing matrix libraries.[4] This matrix library only handles two-dimensional matrices compared to other matrix libraries. In addition, the library lacks object-oriented design, possibly to increase computational efficiency to handle a large volume of training data.
External Links
References
This article "DataPredict" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:DataPredict. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.
- ↑ 1.0 1.1 1.2 Aiman, Aqwam Harish. "DataPredict Documentation". GitHub. Retrieved 28 June 2023.
- ↑ 2.0 2.1 Aiman, Aqwam Harish. "MachineLL Source Codes". GitHub. Retrieved 28 June 2023.
- ↑ 3.0 3.1 "Can Machine Learning and Deep Learning Models Contribute in Roblox Games?". Roblox DevForum. Roblox.
- ↑ 4.0 4.1 Aiman, Aqwam Harish. "MatrixL Documentation". GitHub. Retrieved 28 June 2023.
