Question 3: Will the Robot Function Just as Well When the Data Segmentation, Feature Extraction, and Classification Algorithms are Fully Automated?

As described in other parts of this Wiki, INTEGRAT trains the AI researcher to read the mind of the robot - with regard to examination of individual data examples, feature extraction algorithms, and examination of the training data set as a whole.

One more aspect that can foil even the best trained AI algorithm is the field calibration for use in-situ. Quite a few errors resulting from a change in sensor circuitry can make even an excellent pattern recognition algorithm fail consistently. Besides changes in data range and scaling (which can be compensated for in the field) there may also be changes in the way the data is presented to the algorithm causing a dramatic segmentation change.

Segmentation is how the AI robot chops up a continuous stream of sensor data into discrete chunks which are independently classified. From this "stream of classifications" further, and more refined/abstract/advanced classifications can be determined (such as with deep learning networks, Hidden Markov Models, and linguistic analysis, etc.). If the original laboratory (in vitro) segmentation is very different from the field (in situ) parameters, the algorithms can fail even before the AI is even employed.

Wiki Home Page

Additional details can be found in a patent which describes several novel, non-obvious techniques to examine this issue. See US Patent 5867816:

The AI Robot is not useful if it requires a human expert to accompany the device when it is deployed in the field for day to day use. The robot must work on its own. Nevertheless, a quality assurance mechanism is required that makes the robot demonstrate it is working in “fully automatic” mode while a human expert grades it on the decisions the AI has made.

Independent Claims 1, 11, and 29 – The human expert examines and modifies the segmentation, features, and decision identifications to improve automated functioning.