麻酱

verify-tag4 Million Sudoku Puzzles Easy-to-Hard

gamespuzzlesartificial intelligencebeginnerintermediate

23

已售 0
637.05MB

数据标识:D17171237755845061

发布时间:2024/05/31

以下为卖家选择提供的数据验证报告:

数据描述

Puzzles Vary in Difficulty, Why?

For any Sudoku puzzle, its level of difficulty generally relates to the number of clues it has (i.e. the number of digits on the board to begin with). More clues, easier puzzle. A good solving algorithm should be able to handle very hard puzzles (17 clues minimum to guarantee that it's a reasonable puzzle), while also quickly solve the easiest puzzles (80 clues).

Magic Number 17

Why is 17 the minimum clue number? here is why

This Dataset

Other Sudoku datasets on Kaggle (1, 3, and 9 Million Sudoku Puzzle datasets) failed to address this aspect of the Sudoku game, thus this dataset. This dataset contains 4 Million Sudoku puzzles and solutions, range from 17 clues to 80 clues. There are 62500 puzzles and solutions for each number of clues.

Some Solutions Are Not Unique

Due to mathematical principles, the solutions for some puzzles in this dataset are not unique. You might need an answer checker to handle multiple possible solutions. That being said, comparing your answers with the example solutions in the dataset before using the answer checker will save you a lot of time when running experiments at 10000+ scale.

There is a compatible answer checker available to you in this starter pack

Performance Baseline

A sample implementation of the dancing-link algorithm is in the same starter pack. This is an advanced sudoku solving algorithm and should not be used as a baseline if you are building your own algorithm for fun. But if you are trying to solve sudoku with RL, feel free to compare your model with this sample algorithm on their performances.

data icon
4 Million Sudoku Puzzles Easy-to-Hard
23
已售 0
637.05MB
申请报告