Improving Local Identifiability in Probabilistic Box Embeddings

10/09/2020
by   Shib Sankar Dasgupta, et al.
3

Geometric embeddings have recently received attention for their natural ability to represent transitive asymmetric relations via containment. Box embeddings, where objects are represented by n-dimensional hyperrectangles, are a particularly promising example of such an embedding as they are closed under intersection and their volume can be calculated easily, allowing them to naturally represent calibrated probability distributions. The benefits of geometric embeddings also introduce a problem of local identifiability, however, where whole neighborhoods of parameters result in equivalent loss which impedes learning. Prior work addressed some of these issues by using an approximation to Gaussian convolution over the box parameters, however, this intersection operation also increases the sparsity of the gradient. In this work, we model the box parameters with min and max Gumbel distributions, which were chosen such that space is still closed under the operation of the intersection. The calculation of the expected intersection volume involves all parameters, and we demonstrate experimentally that this drastically improves the ability of such models to learn.

READ FULL TEXT
research
11/27/2019

Information-Geometric Set Embeddings (IGSE): From Sets to Probability Distributions

This letter introduces an abstract learning problem called the “set embe...
research
09/10/2021

Box Embeddings: An open-source library for representation learning using geometric structures

A major factor contributing to the success of modern representation lear...
research
02/28/2022

Description Logic EL++ Embeddings with Intersectional Closure

Many ontologies, in particular in the biomedical domain, are based on th...
research
01/03/2020

Representing graphs as the intersection of cographs and threshold graphs

A graph G is said to be the intersection of graphs G_1,G_2,...,G_k if V(...
research
09/01/2017

A simple en,ex marking rule for degenerate intersection points in 2D polygon clipping

A simple en,ex rule to mark the intersection points of 2D input polygon ...
research
10/04/2019

Maximum Matchings in Geometric Intersection Graphs

Let G be an intersection graph of n geometric objects in the plane. We s...
research
08/23/2023

Recent Developments in Pandora's Box Problem: Variants and Applications

In 1979, Weitzman introduced Pandora's box problem as a framework for se...

Please sign up or login with your details

Forgot password? Click here to reset