oya-ann v0.5.0

OyaAnn is a Javascript library for building, training and using artificial neural networks (ANNs) that approximate and model simple multi-variate relationships. OyaAnn neural networks can be generated dynamically from observational data according to need and serialized for later use.

Consider, for example, temperature compensation for conductivity measurements (i.e., EC/PPM). EC/PPM probes are temperature sensitive and readings will change with temperature. For example, the EC of a reference solution having EC=1413 microsiemens @ 25°C may vary as follows (Atlas Scientific EC-EZO):

To complicate matters further, solutions with different dissolved solids will each exhibit their own individual temperature compensation curves. Modeling, calibrating and measuring conductivity for different solutions at different temperatures is therefore a challenge. What is needed is a simple way to create calibrated models from observed data. Neural networks are ideal for this task.

With OyaAnn, we can generate custom temperature compensation ANNs for new nutrient solutions and train them with locally observed data. Once trained, these ANNs can be archived and re-used as needed.

var examples = [
    new Example([5],[896]),
    new Example([10],[1020]),
    new Example([15],[1147]),
    new Example([20],[1278]),
    new Example([25],[1413]),
    new Example([30],[1548]),
    new Example([35],[1711]),
    new Example([40],[1860]),
    new Example([45],[2009]),
    new Example([50],[2158]),
];
network.train(examples);

Installation

Use npm to install oya-ann.

npm install oya-ann

See Also

• OyaAnn is a divergent fork of Kinann repurposed for OyaMist applications.
• OyaAnn wiki...
• mathjs many thanks to MathJS for expression parsing and derivatives!