Building a Telescope to Look Into High-Dimensional Image Spaces

by   Mitch Hill, et al.

An image pattern can be represented by a probability distribution whose density is concentrated on different low-dimensional subspaces in the high-dimensional image space. Such probability densities have an astronomical number of local modes corresponding to typical pattern appearances. Related groups of modes can join to form macroscopic image basins that represent pattern concepts. Recent works use neural networks that capture high-order image statistics to learn Gibbs models capable of synthesizing realistic images of many patterns. However, characterizing a learned probability density to uncover the Hopfield memories of the model, encoded by the structure of the local modes, remains an open challenge. In this work, we present novel computational experiments that map and visualize the local mode structure of Gibbs densities. Efficient mapping requires identifying the global basins without enumerating the countless modes. Inspired by Grenander's jump-diffusion method, we propose a new MCMC tool called Attraction-Diffusion (AD) that can capture the macroscopic structure of highly non-convex densities by measuring metastability of local modes. AD involves altering the target density with a magnetization potential penalizing distance from a known mode and running an MCMC sample of the altered density to measure the stability of the initial chain state. Using a low-dimensional generator network to facilitate exploration, we map image spaces with up to 12,288 dimensions (64 × 64 pixels in RGB). Our work shows: (1) AD can efficiently map highly non-convex probability densities, (2) metastable regions of pattern probability densities contain coherent groups of images, and (3) the perceptibility of differences between training images influences the metastability of image basins.


page 6

page 11

page 27

page 28

page 30

page 31

page 32

page 33


High-Dimensional Probability Estimation with Deep Density Models

One of the fundamental problems in machine learning is the estimation of...

Continuous Herded Gibbs Sampling

Herding is a technique to sequentially generate deterministic samples fr...

Preserving local densities in low-dimensional embeddings

Low-dimensional embeddings and visualizations are an indispensable tool ...

Density Estimation with Distribution Element Trees

The estimation of probability densities based on available data is a cen...

A `periodic table' of modes and maximum a posteriori estimators

The last decade has seen many attempts to generalise the definition of m...

Adaptive Cluster Expansion (ACE): A Hierarchical Bayesian Network

Using the maximum entropy method, we derive the "adaptive cluster expans...

Transporting Densities Across Dimensions

Even the best scientific equipment can only partially observe reality. R...

Please sign up or login with your details

Forgot password? Click here to reset