I am pleased to announce the 1.1.0 release of Project Zenith. The main focus of this release was to improve performance (in terms of speed). It boasts a decrease in the time it takes to complete one epoch from 20 seconds down to 500 milliseconds (a 40 times improvement). In simpler terms, it now only takes half a second to solve the z408, with a probability of success of 87% using the default configuration.

The probability of success has also increased significantly due to using a language model built on a different corpus (The Blog Authorship Corpus). I think this corpus more closely models the language style used by the zodiac in particular.

With respect to the z408, here are some quick stats on the success rate and time per epoch, based on the number of iterations through the set of ciphertext keys per epoch:
– when choosing 2500 iterations, each epoch takes 250 ms, and 680 out of 1000 epochs (68.00%) produced the correct solution.
– when choosing 5000 iterations, each epoch takes 500 ms, and 873 out of 1000 epochs (87.30%) produced the correct solution.
– when choosing 10000 iterations, each epoch takes 1 second, and 930 out of 1000 epochs (93.00%) produced the correct solution.
– when choosing 20000 iterations, each epoch takes 2 seconds, and 985 out of 1000 epochs (98.50%) produced the correct solution.

So quite a high probability of success can be achieved by increasing the number of iterations, with the effect being a slower time to completion per epoch.

Some additional transformers have been added as well, but I’ll save those details for the README. I hope this release increases the usefulness/utility of the project. Please share any questions or feedback.

https://github.com/beldenge/Zenith/releases/tag/1.1.0