The Ekkono SDK is filled with things that will make it easy to develop high end machine learning solutions. Here are some of the key features of our SDK.
Ekkono helps your products get smarter the more you use them. We use incremental learning to improve the algorithms with every new sample of data the you feed them. This enables each machine to adapt to the exact conditions that it is operating under.
With conformal prediction we can give your model belts and suspenders. We supply a predicted value of the target sensor, but we also give you a probability range. This enables statements like: within 99.99 per cent certainty, this sensor will not reach a critical level within the next hour.
Change and anomaly detection
Ideally you have data from thousands of failures of your product. Then we can analyse this data and detect future crashes. In reality, this failure data doesn’t exist.
We turn this around. We analyze how the product behaves when healthy. When it starts to fail, our change and anomaly detectors swiftly finds the error before the product breaks down.
High speed data is difficult to handle. We can help you find some order in the chaos by our Spectral library. Spectral includes functionality for Fast Fourier Transform (FFT) and Autocorrelation as well as Wavelet Decomposition and Signal to Noise Ratio.