Neuton (network)
| Artificial intelligence |
|---|
| Major goals |
| Approaches |
| Philosophy |
| History |
| Technology |
| Glossary |
Neuton is a neural network framework and automated machine learning (AutoML) solution developed by Bell Integrator. It is based on its own proprietary algorithm protected by a patent issued in 2019.
Overview[edit]
Neuton allows users to build, train and deploy neural network models to solve problems of multivariate regression and classification. Neuton is available as an enterprise solution to run on-premises or an online service on Google Cloud Platform. Predictions are made via a REST API or Web interface. The resulting model can be used as a service deployed in the cloud or downloaded for local use.[1] Benchmark tests verify that Neuton outperforms competitors by several orders of magnitude.[2]
Features[edit]
Neural Network Framework[edit]
- Automated Neuton model architecture creation[3]
- Automated hyper parameters configuration[4]
- Regression and Classification problem solving
- No limitations on dataset size for training
- No overfitting
- Automated training completion upon reaching optimal model and predictive power while avoiding overfitting[5]
Auto ML[edit]
- Automatic and manual problem type definition (regression or classification) based on dataset analysis
- Automated dataset preparation (preprocessing)
- Automated feature engineering
- Automated training
- Automatic training completion upon user input thresholds reach[6]
- CSV datasets support
Preloaded datasets
- Models quality metrics visualization.
Predictions[edit]
- Capability to download trained models in HDF5 format with Python API for local usage along with script examples on how to use Neuton models offline
- Web Interface for predictions on trained models
- Predictions via REST API
- No limitations on input data size for predictions[7]
Infrastructure[edit]
- Cloud Based solution
- SaaS solution
- GPU Support for training[8]
- Automatic and seamless provisioning and release of infrastructure necessary to perform training and web hosting for predictions[9]
References[edit]
- ↑ "Neuton: another, problematic neural system structure". October 5, 2018.
- ↑ Integrator, Bell (October 4, 2018). "Bell Integrator Launches Neural Network Framework to democratize AI". GlobeNewswire News Room.
- ↑ "Bell Integrator запускает прорывную нейронную сеть!". SaratovIT. October 10, 2018.
- ↑ Cheung, K. C. (October 9, 2018). "Neuton: A New Disruptive Neural Network Framework Far More Effective Than Any Other Framework".
- ↑ "Bell Integrator Launches Disruptive Neural Network Framework". bell.one.
- ↑ "NEUTON Trademark of Bell Integrator, Inc. Serial Number: 87896515 :: Trademarkia Trademarks". trademark.trademarkia.com.
- ↑ "Фреймворк Neuton сделает искусственный интеллект доступным непрофессионалам". itWeek.
- ↑ Anadiotis, George. "Neuton: A new, disruptive neural network framework for AI applications". ZDNet.
- ↑ "Auto ML solution makes AI 'available to everyone'". Smart2.0. October 11, 2018.
Further reading[edit]
External links[edit]
This article "Neuton (network)" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:Neuton (network). Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.
