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(,), new Example(,), new Example(,), new Example(,), new Example(,), new Example(,), new Example(,), new Example(,), new Example(,), new Example(,), ]; network.train(examples);
npm to install oya-ann.
npm install oya-ann