No Model Size to
Quality Trade Off
Our unique framework allows creation of a neural network structure of optimal size & accuracy.
Neuton’s models are extremely compact:
Up to 1000 times
Fewer coefficients and neurons
Smaller in size (Kb)
Faster inference
in comparison to TensorFlow and other frameworks and algorithms.
Bring Intelligence to the Edge!
A comprehensive approach to building exceptionally tiny and explainable models for Data Scientists and non-Data Scientists alike - opens up amazing possibilities for smart devices to do complex computing on the edge. We are solving tasks for regression, time series, classification with tabular data. Support for audio, image and video is on the roadmap.
Automation
Neuton is a zero-code SaaS based AutoML platform developed for All
Unique Neural
Network Framework
Neuton’s disruptive neural network framework builds extremely compact models
Explainability
Neuton Explainability Office offers a unique set of tools for model quality evaluation and understanding the logic behind the model
Learn How Neuton Creates Incredibly Compact and Accurate Models
Selective connections
Unique algorithm
Automatic structure growth
No manual search
Constant cross-validation
Selective approach to connected features
First, Neuton uses a selective approach to connected features. Neuton only needs to make connections with the most important features
Unique patented global optimization algorithm
We leverage a unique patented algorithm to adjust the coefficients within a model. Unlike most algorithms Neuton’s algorithm is not based on backward propagation of errors and the stochastic gradient descent. Furthermore, the Neuton algorithm minimizes the likelihood of hitting local minima. This is especially important when building small models where the probability of hitting local minima is particularly high. Our algorithmic approach helps to avoid this issue.
Automatic neuron-by-neuron network structure growth
The Neuton platform enables models to be built automatically, neuron by neuron, starting from learning general features and moving toward identifying the most specific ones. This allows selection of a model of almost any level of precision and size, in a single iteration.
No manual search for neural network parameters
Neuton eliminates the time-consuming multidimensional manual search for neural network parameters (number of layers, neurons in a layer, type of activation function, batch size, learning rate, etc.) and allows one to quickly & efficiently find the optimal structure.
Constant cross-validation
The step-by-step growth of the neural network makes it possible to cross-validate with the addition of each neuron, which is typically not feasible with a standard approach. Furthermore, constant cross-validation increases the generalizing capabilities of the model, which allows for creation of compact models - without compromising accuracy.
Build an Efficient Self-Organizing Neural Network
Neuton’s neural network learning is not based on backward propagation of errors and the stochastic gradient descent algorithm, but rather uses a new efficient global optimization algorithm with an excellent generalization capability, allowing for development of the optimal network structure for one iteration.
Evaluate the Uniqueness of our Approach by Comparing our Benchmarks
For example, Neuton’s Model for Combined Cycle Power Plant Data Set is 208.6 times smaller than TensorFlow, and 42.3 times smaller than TensorFlow Lite.
Algorithm
Size, kB
Coefficients
Target metric
Holdout metric
TensorFlow
31.08
2 338
MAE
3.35
TensorFlow Lite
6.31
n/a
MAE
3.36
Neuton
0.149
100
MAE
3.30
Combined Cycle Power Plant Data Set 
The dataset contains 9,568 points collected from a Combined Cycle Power Plant over 6 years. Features consist of hourly average ambient variables (Temperature, Ambient Pressure, Relative Humidity and Exhaust Vacuum) used to predict the net hourly electrical energy output.
Create Tiny Models without Сompression
Neuton models maintain all of their original characteristics, without any reduction of accuracy. Neuton does not reduce the model size after its creation.
Neuton does not use quantization, pruning, clustering, nor distillation.
Embed into Edge Devices
Neuton’s models can be built into microcontrollers and into other small compute devices, with limits as challenging as the following characteristics:
Energy - 10s-100s mAh
Processor < 100 MHz
Memory < 100 Kb
Explore the Case Study Determining Cardiac Arrhythmias
Learn how to determine cardiac arrhythmias on a microcontroller without the cloud fast and easily leveraging the Neuton AutoML platform.
Improve Quality of your Models Ensembling them with Neuton Models
Thanks to unique model creation approach, models generated by the Neuton Neuron Network Framework enhance any model in the ensemble
Benchmarks
To benchmark a model correctly, and allow for a clear comparison against other solutions, Neuton has three measurements: number of coefficients, model size, and Kaggle score.
TinyML cases
Well-known Kaggle cases
News
Features
Contact Us
Get in touch to learn more
Contact Us
Contact Us
Thank you!
Please check your inbox. We will reach out to you shortly with the next steps.
Contact Us
Thank you!
Please check your inbox. We will reach out to you shortly with the next steps.