Distributional encoding represents each level of a categorical input as the empirical distribution of the output at that level, and uses kernels between distributions inside a Gaussian process.
We plan to investigate further its potential on large datasets in future work
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Distributional encoding for Gaussian process regression with qualitative inputs
Distributional encoding represents each level of a categorical input as the empirical distribution of the output at that level, and uses kernels between distributions inside a Gaussian process.