We propose two crossword generation methods based on a crossword concatenation, word addition and crossword rotation operation. This can be viewed as an alternative to the method proposed by Bonomo, Lauf and Yampolskiy or Bulitko and Botea, who focus on generating matrices filled with letters and mutating them in order to make them into actual crosswords. The first one uses a combination of first improvement and best improvement local search methods. The choice on which one to use is made using the temperature calculated for a given turn. Second algorithm is a simulated annealing algorithm which uses best improvement search and word removal operation. The crosswords are evaluated using a goal function that includes the amount of intersections in a crossword and the density of letters in the crossword. Unfortunately, both of these solutions, while producing decent results, create puzzles unsolvable for humans in reasonable time. Because of that, we plan on implementing: a better goal function, targeted word removal and targeted word addition. We also plan to switch simulated annealing for a cuckoo search algorithm.

The class of bivariate Archimax copulas spans a broad range of dependence. However, its size is much reduced when the degree of association is prespecified. The extent of this reduction is studied for a large set of measures of association. The cases of exchangeable and non-exchangeable Archimax copulas are both considered.