vicdoja/metacountdown

A Python package to solve Countdown's number round using metaheuristics.

metacountdown's Issues

Encapsulate functions inside classes

This issue refers to the encapsulation of function inside classes, insead of having everything inside a "utils" module.

Bug within the linear tree representation

The evaluation of the linear tree is acting strange, and I just noticed.

Apparently small expressions are giving optimal fitness for relatively big objectives.

This needs to be looked into ASAP.

Write a useful README

The README file is one of the most important files in a project.

This is like this because the README file is the first content that users or other developers see when they first enter the project page.

For our early development stage, it should include the following sections:

• Description: define the problem, it complexity and the used emetaheuristics.
• Origin: this is a Master's subject project.
• Contributing
• References: include links to revelant materials to the project
• License

Implement code for generating useful graphs

This will be very useful for the writing of the presentation, and might be a massive time-saver.

• Best-Fitness/Generation

• Avg-Fitness/Generation

• % of valid individuals

• Simulated annealing graphics

PS: More useful graphs can be added after the creation of this issue.

Document EVERYTHING

I think that this issue doesn't need further explanation. The only notable comment here is that the docstrings will be written using the Google style.

Improve testing

This includes implementing a mechanism to store problem instances, their best solution up to now and the results of using genetic algorithms and simulated annealing.

• Problem instances
• Best solutions
• Tables of results

Heal or penalize invalid individuals

As the course's professor told me: when an individual suffers from a condition 🤒, it may heal 💉, get handicapped 🤕 or die ☠.

I don't like the dying approach, so don't get me near any of that stuff. In our case, every individual has the potential to become the fittest one 💪, so we must try to either heal 🏥 them or penalize 🎃 their fitness.

Note that this may help with the covering issues of the linear tree ➗.

Implement algorithm loop (or several)

Include algorithm loops inside the library, such that its use can be parametrized.

Complete alternative individual codification/operators

This task involves finishing the implementation of the alternative individual codification and operators (generate_tree_alt, mutate_tree_alt, mate_tree_alt).

This representation involves a tuple of two lists. The first list is made of N terminals (numbers) of the evaluation tree, and the second list is made of N-1 non-terminal nodes (mathematical operators) that specify the operations inside the evaluation tree.

• generate_tree_alt
• mutate_tree_alt
• mate_tree_alt