Automated Algorithm Selection: from Feature-Based to Feature-Free Approaches

Abstract

We propose a novel technique for algorithm-selection, applicable to optimisation domains in which there is implicit sequential information encapsulated in the data, e.g., in online bin-packing. Specifically we train two types of recurrent neural networks to predict a packing heuristic in online bin-packing, selecting from four well-known heuristics. As input, the RNN methods only use the sequence of item-sizes. This contrasts to typical approaches to algorithm-selection which require a model to be trained using domain-specific instance features that need to be first derived from the input data. The RNN approaches are shown to be capable of achieving within 5% of the oracle performance on between 80.88 and 97.63% of the instances, depending on the dataset. They are also shown to outperform classical machine learning models trained using derived features. Finally, we hypothesise that the proposed methods perform well when the instances exhibit some implicit structure that results in discriminatory performance with respect to a set of heuristics. We test this hypothesis by generating fourteen new datasets with increasing levels of structure, and show that there is a critical threshold of structure required before algorithm-selection delivers benefit.

Generating diverse and discriminatory knapsack instances by searching for novelty in variable dimensions of feature-space

“…Finally, we note that all algorithms were written in C++. 3 The experimental assessment was conducted to address two main questions using data obtained from six different experiments: three experiments using a high-dimensional feature vector (8D, 6D, 4D) to calculate novelty using 𝑁 𝑆 𝑓 , and three experiments using the 2D projection obtained from the corresponding high-dimensional model to calculate novelty using 𝑁 𝑆 𝑃𝐶𝐴 . Particularly, we address the following research questions:…”

“…We leave this for further work. Future work will be directed towards evaluating the method in additional domains, for example in TSP, where there has already been some interest in generating diverse instances [4,5]; and in binpacking, where there has also been some preliminary work in trying to produce diverse instances [3]. As noted in Section 6, an obvious avenue for future investigation is also to consider different methods of learning new descriptors.…”

Revisit the Algorithm Selection Problem for TSP with Spatial Information Enhanced Graph Neural Networks

“…In [21], researchers sample landscape information from instances and transform it into images, then apply CNN to select algorithms for Black-Box Optimization Benchmarking (BBOB) function instances. The authors of [22] treat online 1D Bin-Packing Problem instances as sequence data and apply Long Short-Term Memory (LSTM) to predict heuristic algorithms' performance. In the feature-free algorithm selection field, instances are usually converted to images or sequences, and graph representations are seldom used.…”