- Optimal back-to-front airplane boarding
The problem of finding an optimal back-to-front
airplane boarding policy is explored, using the mathematical
model introduced by Bachmat et al. A combined
analytical and numerical solution is presented
for the number of passenger groups m=2.
For a larger m, a similarly constructed recursive
computation yielding an optimal policy is described.
Applying it numerically,
optimal policies for m=3,4 are found;
modern transport aircraft design
renders further increase of m impractical.
- FORTH stack optimization
- An exhaustive searcher for the best source-level optimization
of stack manipulation sequences encountered in an input
FORTH code snippet.
Can be used as a learning aid to study the basic stack accessor words
in FORTH. Written in Java. (Undergraduate project, based on
my joint final work with Ram Janovski for the
Functional Programming 2004
- Comparing dbparser and MINIPAR
- Two probabilistic parsers, dbparser and MINIPAR, are
on a subset of
the WSJ corpus from the Penn Treebank
according to the tree accuracy criterion.
The sentences selected were the most difficult ones
in the corpus,
according to three empirically chosen criteria ---
parse tree width, height, and the top-level verb ambiguity.
The two parsers perform similarly on the first two groups,
while on the third one dbparser is significantly better.
Further possible comparison directions are then outlined.
(Final work in the Topics in natural language processing 2005 course).