
Model Inversion Networks for ModelBased Optimization
In this work, we aim to solve datadriven optimization problems, where t...
read it

Amortized Conditional Normalized Maximum Likelihood
While deep neural networks provide good performance for a range of chall...
read it

Information criteria for nonnormalized models
Many statistical models are given in the form of nonnormalized densitie...
read it

A statistical theory of outofdistribution detection
We introduce a principled approach to detecting outofdistribution (OOD...
read it

HighDimensional Probability Estimation with Deep Density Models
One of the fundamental problems in machine learning is the estimation of...
read it

Goaldirected Generation of Discrete Structures with Conditional Generative Models
Despite recent advances, goaldirected generation of structured discrete...
read it

Mean Reverting Portfolios via Penalized OULikelihood Estimation
We study an optimizationbased approach to con struct a meanreverting ...
read it
Offline ModelBased Optimization via Normalized Maximum Likelihood Estimation
In this work we consider datadriven optimization problems where one must maximize a function given only queries at a fixed set of points. This problem setting emerges in many domains where function evaluation is a complex and expensive process, such as in the design of materials, vehicles, or neural network architectures. Because the available data typically only covers a small manifold of the possible space of inputs, a principal challenge is to be able to construct algorithms that can reason about uncertainty and outofdistribution values, since a naive optimizer can easily exploit an estimated model to return adversarial inputs. We propose to tackle this problem by leveraging the normalized maximumlikelihood (NML) estimator, which provides a principled approach to handling uncertainty and outofdistribution inputs. While in the standard formulation NML is intractable, we propose a tractable approximation that allows us to scale our method to highcapacity neural network models. We demonstrate that our method can effectively optimize highdimensional design problems in a variety of disciplines such as chemistry, biology, and materials engineering.
READ FULL TEXT
Comments
There are no comments yet.