50%(2)1 out of 2 people found this document helpful
This preview shows page 1 - 3 out of 4 pages.
1Project 1 ReflectionSukeerthi Varadarajan[email protected]Question 1How does the agent work?The agent solves the problem by representing the objects as sentences in aSemantic Network. We then apply Generate and Test to compare each of theanswer options as figure D and then comparing the transformations andalignments between figures A, B, C and D, i.e. A to B, C to D and so on. Then theagent compares the pairs thus created to arrive at the answer. As there are sixpotential answers to the question, we use similarity scores to pick the bestmatch. The greater the similarity, the better the match.Question 2What was your approach towards solving the problem?Initially we started by solving the basic problems B-01 and B-02, which are easierto solve. Once we created the agent to solve the easier problems, we thentweaked the agent to answer further question. As the problems progressed,there are more transformations happening to figure D, so we added differenttransformation weights to help the agent identify the correct answer from thegiven options.Figure 1.Increasing complexity of the Basic Problems
2The agent uses similarity metrics to score each of the options. The option withthe highest similarity score is chosen as the answer. The transformations arerecorded and stored in thetransformsdictionary, along with the similarity scoresfor each of the correct transformations.