• Corpus ID: 238419059

Multi-objective Optimization by Learning Space Partitions

  title={Multi-objective Optimization by Learning Space Partitions},
  author={Yiyang Zhao and Linnan Wang and Kevin Yang and Tianjun Zhang and Tian Guo and Yuandong Tian},
In contrast to single-objective optimization (SOO), multi-objective optimization (MOO) requires an optimizer to find the Pareto frontier, a subset of feasible solutions that are not dominated by other feasible solutions. In this paper, we propose LaMOO, a novel multi-objective optimizer that learns a model from observed samples to partition the search space and then focus on promising regions that are likely to contain a subset of the Pareto frontier. The partitioning is based on the dominance… 

Three-Way Trade-Off in Multi-Objective Learning: Optimization, Generalization and Conflict-Avoidance

This work focuses on a new stochastic variant of MGDA - the Multi-objective gradient with Double sampling (MoDo) algorithm, and studies the generalization performance of the dynamic weighting-based MoDo and its interplay with optimization through the lens of algorithm stability.

Assessing multi-objective optimization of molecules with genetic algorithms against relevant baselines

It is shown that both CHIMERA and Hypervolume achieve better formal optimality than the baselines and generate molecules closer to a user-specified Utopian point in property space, mimicking typical materials design objectives.

Pareto Set Learning for Expensive Multi-Objective Optimization

A novel learning-based method to approximate the whole Pareto set for MOBO, which generalizes the decomposition-based multi-objective optimization algorithm (MOEA/D) from finite populations to models and design a simple and powerful acquisition search method based on the learned Pare to set, which naturally supports batch evaluation.

Landscape Optimization for Prescribed Burns in Wildfire Mitigation Planning

Wildfires have increased in extent and severity, and are posing a growing threat to people’s well-being and the environment. Prescribed burns (burning on purpose parts of the landscape) are one of

Multi-Objective Bayesian Optimization over High-Dimensional Search Spaces

MORBO significantly advances the state-of-the-art in sample efficiency for several high-dimensional synthetic problems and real world applications, including an optical display design problem and a vehicle design problem with 146 and 222 parameters, respectively.

FuncPipe: A Pipelined Serverless Framework for Fast and Cost-Efficient Training of Deep Learning Models

FuncPipe is designed with the key insight that model partitioning can be leveraged to bridge both memory and bandwidth gaps between the capacity of serverless functions and the requirement of DL training, and it achieves 7%-77% cost savings and 1.3X-2X speedup compared to state-of-the-art serverless-based frameworks.

Multiobjective optimization for crash safety design of vehicles using stepwise regression model

A nondominated sorting genetic algorithm is employed to search for Pareto solution to a full-scale vehicle design problem that undergoes both the full frontal and 40% offset-frontal crashes, demonstrating the capability and potential of this procedure in solving the crashworthiness design of vehicles.

Differentiable Expected Hypervolume Improvement for Parallel Multi-Objective Bayesian Optimization

This work derives a novel formulation of Expected Hypervolume Improvement, an acquisition function that extends EHVI to the parallel, constrained evaluation setting and demonstrates that it is computationally tractable in many practical scenarios and outperforms state-of-the-art multi-objective BO algorithms at a fraction of their wall time.

NAS-Bench-201: Extending the Scope of Reproducible Neural Architecture Search

This work proposes an extension to NAS-bench-101: NAS-Bench-201 with a different search space, results on multiple datasets, and more diagnostic information, which provides additional diagnostic information such as fine-grained loss and accuracy, which can give inspirations to new designs of NAS algorithms.

Max-value Entropy Search for Multi-Objective Bayesian Optimization

This work proposes a novel approach referred to as Max-value Entropy Search for Multi-objective Optimization (MESMO), which employs an output-space entropy based acquisition function to efficiently select the sequence of inputs for evaluation for quickly uncovering high-quality solutions.

Covariance Matrix Adaptation for Multi-objective Optimization

A single-objective, elitist, CMA-ES is introduced using plus-selection and step size control based on a success rule and a population of individuals that adapt their search strategy as in the elitists is maintained, subject to multi-objectives selection.

Steady-State Selection and Efficient Covariance Matrix Update in the Multi-objective CMA-ES

A generational and two steady-state selection schemes for the MO-CMA-ES, a multi-objective covariance matrix adaptation evolution strategy that combines a mutation operator that adapts its search distribution to the underlying optimization problem with multicriteria selection, are compared.

Learning Search Space Partition for Black-box Optimization using Monte Carlo Tree Search

LA-MCTS serves as a meta-algorithm by using existing black- box optimizers as its local models, achieving strong performance in general black-box optimization and reinforcement learning benchmarks, in particular for high-dimensional problems.

A Flexible Multi-Objective Bayesian Optimization Approach using Random Scalarizations

This work proposes an approach based on random scalarizations of the objectives that can focus its sampling on certain regions of the Pareto front while being flexible enough to sample from the entire Pare to front if required, and is less computationally demanding compared to other existing approaches.

SMS-EMOA: Multiobjective selection based on dominated hypervolume

Multiobjective Optimization on a Limited Budget of Evaluations Using Model-Assisted -Metric Selection

This paper provides a review of contemporary multiobjective approaches based on the singleobjective meta-model-assisted 'Efficient Global Optimization' (EGO) procedure and describes their main concepts and introduces a new EGO-based MOOA, which utilizes the $\mathcal{S}$-metric or hypervolume contribution to decide which solution is evaluated next.