yacinemahdid/ai-boggle-solver

boggle solver powered by artificial intelligence! ⚡

Hi there 👋 I'm Yacine!

I'm the CTO of GRAD4 a tech company which enable manufacturing companies to connect, find and collaborate with local supplier 10 times faster than their current process!

I'm also a PhD student at the Biosignal Interaction and Personhood Technology Lab (BIAPT) lab at McGill university using machine learning techniques to gain insight into the brain correlate of consciousness and pain!

I love to teach machine learning and computational neuroscience, you can head over to my Youtube channel to check out my tutorials!

The best way to reach me out for business inquiry is through Linkedin, or Twitter for less pressing issues!

ai-boggle-solver's Issues

bug adjacency matrix isn't properly setup creating connection through all nodes in the graph

The way it is currently setup for a board like this:

n,y,r
n,o,a
o,r,s

is like that:

True, True, True
True, True, True
True, True, True

This isn't a proper adjacency graph, even if we constrain how the words can be created. Either we need to rename this variable to just a is_visited_matrix or we need to change this into an actual adjacency matrix.

Complete the README.md to match with what the end project should be able to do!

Improve the Python algorithm so that it makes use of adjacency matrix

This is the first step in order to gate which letter is connected to which other letter!

• 🎥 Explain the usage of the Adjacency graph
• 🎥 Show the difference when we add an adjacency matrix to the code
• 🎥 Show the improvement enabled by the graph to cut down some connection that are impossible
• Merge this MR

Create the folder structure so that the project as a solid base for the data science part

I'll use this structure and modify has needed: https://drivendata.github.io/cookiecutter-data-science/

Document the algorithm that needs to be implemented in Rust in order to find the right words the fastest

Create a rough script for the first iteration

I really need to have a script when going through the algorithm as I usually get pretty side-tracked in non essential point!

Document how to implement the OpenCV part to segment and detine the boundary of the boggle dices

Find a base model that can do digit recognition on images

Setup the input and output for the whole project in Python

Inputs

In terms of input we are talking about a grid like this:

[] [] [] [] []
[] [] [] [] []
[] [] [] [] []
[] [] [] [] []
[] [] [] [] []

with letters in there!

Output

In terms of output we are looking at this:

• word-1
• word-2
• word 3
• etc.

Conclusion

I'll setup the script for "reading" input and to output a proper list of words when calling the main function

Setup your virtual environment workflow

Document in milestone the stages of development and the deliverables

I'll aim for a progressive always working a bit better milestone setup.

Create a lucid chart diagram of what I think all of the interconnected component look like for the doc

Change the API so that we can input different sort of dictionary file

    # find all the words in the grid
    # TODO REMOVE THIS HARD CODED DICTIONARY
    words = find_all_words(grid, 'en_dict.txt')

right now we have the above, but this is hard coded inside the code. Could be easily added into the API

Improve the Python algorithm to include dynamic programming with a english language decision tree

The idea is to encode the whole dictionary pre-boggle game into a decision tree where each leaf is a word that is valid. While doing the search on the boggle game, we would always early stop when the path we are going into isn't leading to anymore words.

This will ensure optimal search because everything is properly structured beforehand to be able to do the lookup in O(1) and without any extra wasteful work!

Figure out how to structure the part that uses Rust