Ph.D. Thesis: Wizard: Learning Macro-Actions Comprehensively for Planning. Wizard has several desirable properties: it acquires macros for planners and domains but without exploiting any of their specific properties; it captures macros from given example plans to encapsulate planners' experiences on domain landscapes and; further, it evolves them to explore other macros that are not observable from the plans or are normally excluded by existing methods; furthermore, it explores macro-sets comprising macros that could interact among themselves to maximise benefits over the overheads; moreover, it acquires macros from plans of smaller problems and evaluates them against small problems with a view to applying them in larger problems. Overall, Wizard's architecture represents a comprehensive learning framework while its implementation presents a readily available macro learning mechanism that works with arbitrary planners and domains. For further details on Wizard and related publications, please click here.
B.Sc.Engg. Thesis: An Efficient Algorithm for Optimal Vertex-Ranking of Permutation Graphs. A vertex-ranking of a graph G = (V,E) is a labeling of the vertices of G such that every path between any two vertices with the same label contains a vertex with a larger label. A vertex-ranking is optimal if the number of labels used is as small as possible. A graph G is a permutation graph if V = {1, 2, . . ., n} and E = {(i, j) | (i - j)*(p-1(i) - p-1(j)) < 0}, where p = [p(1), p(2), . . ., p(n)] is a permutation of the numbers 1, 2, . . ., n and p-1(i) is the position of i in the permutation. This algorithm finds optimal vertex-rankings of permutation graphs in time O(n3) whereas existing algorithms take O(n6) time.
M.Sc.Engg. Thesis: A parallel algorithm for Optimal Vertex-Ranking of Permutation Graphs. This is a parallel version of the previous O(n3) sequential algorithm. The parallel algorithm runs in O(n3logn) time on Concurrent Read Exclusive Write (CREW) Parallel Random Access Machine (PRAM).