Regularization is all you need: simple neural nets can excel on tabular data

Re:Regularization is all you need: simple neural nets can ex... by lurkmurk on Tuesday June 22, 2021 @07:09PM (#27598497)

They don't? One is a row in table (R^n) (order matters) and the other is a set of points (R^nxd).

If you consider the whole table, any dataSET is like point clouds.

[ Parent | on HN ]

Re:Regularization is all you need: simple neural nets can ex... by whatshisface on Tuesday June 22, 2021 @07:16PM (#27598572)

The table is the point cloud the row is the point. The symmetry is permutations of rows or permutations of points.

[ Parent | on HN ]

Re:Regularization is all you need: simple neural nets can ex... by lurkmurk on Tuesday June 22, 2021 @07:27PM (#27598693)

In which case would you use the whole table?

[ Parent | on HN ]

Re:Regularization is all you need: simple neural nets can ex... by whatshisface on Tuesday June 22, 2021 @08:13PM (#27599101)

Well yeah, there's no permutation symmetry between columns. (Unless there is...)

[ Parent | on HN ]

Re:Regularization is all you need: simple neural nets can ex... by elcomet on Wednesday June 23, 2021 @02:25AM (#27601402)

Then as parent said, almost all datasets are like point cloud, except time-series datasets.

[ Parent | on HN ]

Re:Regularization is all you need: simple neural nets can ex... by whatshisface on Wednesday June 23, 2021 @10:13AM (#27604778)

Yeah, they are.

[ Parent | on HN ]