Our Technology

Home — Our Services — Our Technology — Past Projects — Demos — About Us — Contact Us

Predictive Technology

Confidence Machine is the next generation of machine learning framework based on a breakthrough mathematical theory of algorithmic randomness.

As well as making predictions with the same performance observed in best machine learning algorithms (such as SVM), Confidence Machine makes hedged predictions - predictions with quantitative measures of accuracy and reliability of predictions for high-dimensional data. This new approach, of course, makes significant improvement in the decision-making.

Benefits of Our Technology

The patented Intellectual Properties deliver "hedged" predictions: predictions with valid measures of confidence for high dimensional problems in pattern recognition and regression estimation problems.

Universal Prediction	Traditional Statistical Techniques
Valid measures of confidence of prediction: conformal predictors never overrate accuracy and reliability of their predictions.	Often the property of validity does not hold, and the results are misleading if underlying assumptions are violated.
Generate a measure of confidence for each individual example.	Confidence is given usually for estimation of parameters.
Have a wider variety of applications since required only one (iid) assumption.	Often use restrictive assumptions (like normality).
Developed for both high and low dimensional data.	Usually applied for low dimensional problems.
Allows to control (up to statistical fluctuations) a number of erroneous predictions by selecting a suitable confidence level.	No control of accuracy of predictions once all parameters have been chosen.