Within out iML experiments, everytime we want to present the user with a new choice to provide feedback to the algorithm, we need to have relevant examples at hand. This presents us with an inverted-mindset-problem, as the implemented algorithm is iteratively building up new clusters, at each point choosing the best candidate with respect to a specific cost function. In our iML experiments on the other hand, we will need to ask the user to decide to which cluster a specific datapoint should be added, considering the information loss with each specific choice.. In order to accomplish this we will have to:
- Re-design the algorithm so that it builds clusters "in parallel" instead of "in sequence"
- Write test cases for this new algorithm
- Implement and..
- Test if the results of this algorithm are the same (or comparable in performance) with the original
- Implement an API that tells the algorithm how often to pause and wait for a user input..
Within out iML experiments, everytime we want to present the user with a new choice to provide feedback to the algorithm, we need to have relevant examples at hand. This presents us with an inverted-mindset-problem, as the implemented algorithm is iteratively building up new clusters, at each point choosing the best candidate with respect to a specific cost function. In our iML experiments on the other hand, we will need to ask the user to decide to which cluster a specific datapoint should be added, considering the information loss with each specific choice.. In order to accomplish this we will have to: