GroundSeer™ Regressor: Step 1.

Welcome to the GroundSeer™ Regressor Instant Proof of Concept (POC) App

Here you can test if our ground truth inference algorithms can help improve your ensemble systems, reduce errors in production data pipelines, pick best of breed ML algoritms, or reduce annotation costs.

Ground Truth Inference algorithms are very much like Data Streaming ones. They calculate a statistic of ground truth for your data. One advantage of this is that we can offer our SaaS service while you retain the privacy of your data. We don't want to see your data! And we don't need to see it to infer statistics about its ground truth.

  1. Download this Python code, calculate the deltas for your regressors and save the output to a file.
  2. Submit the output file from step 1 to measure your regressors' relative bias.

If you don't have a dataset, you can manufacture one with this synthetic dataset Python code.

Relative Regressor Bias is the 1st piece of knowledge you need to auto-focus your regressors. If their average predictions are shifted relative to each other, you need to correct for that. Our innovation is that among the infinite ways to do that correction, we pick the one that is engineering error correcting - the smallest (l_1 minimizing) bias correction that will align them.

NOTE: Currently we can only handle 9 regressors in Step 2 below, but between 6 and 9 for Step 1.