inst/gallery/Connect4/README.md

Data

The graphs in this directory are from a dataset of game states of connect 4. Each game state is classified as being a win for player 1, a win for player 2, or a tie. The dataset can be found at dataset link The first 42 columns represent the piece located at a point on the board and the 43rd column represents the player who would go on to win the game. There are 43 columns and 376,430 rows of data.

Interpretation

prVis seems to have a pattern that was not present in umap or tsne. As datapoints stray further from the main cluster on the righthand side, games seem more likely to end up being a tie. It seems like games that are likely to end up in a tie could have a similar structure that is evident even when viewed only with two principal components

Graphs

prVis graphs

prVis without labels

prVis with labels

tsne graphs

tsne without labels

tsne with labels

umap graphs

umap without labels

umap with labels



matloff/prVis documentation built on Dec. 3, 2019, 9:27 a.m.