Darrell Whitley

Tim Whitaker, Darrell Whitley

Neuroevolution is a promising area of research that combines evolutionary algorithms with neural networks. A popular subclass of neuroevolutionary methods, called evolution strategies, relies on dense noise perturbations to mutate networks, which can be sample inefficient and challenging for large models with millions of parameters. We introduce an approach to alleviating this problem by decomposing dense mutations into low-dimensional subspaces. Restricting mutations in this way can significantly reduce variance as networks can handle stronger perturbations while maintaining performance, which enables a more controlled and targeted evolution of deep networks. This approach is uniquely effective for the task of fine tuning pre-trained models, which is an increasingly valuable area of research as networks continue to scale in size and open source models become more widely available. Furthermore, we show how this work naturally connects to ensemble learning where sparse mutations encourage diversity among children such that their combined predictions can reliably improve performance. We conduct the first large scale exploration of neuroevolutionary fine tuning and ensembling on the notoriously difficult ImageNet dataset, where we see small generalization improvements with only a single evolutionary generation using nearly a dozen different deep neural network architectures.

Matt Gorbett, Darrell Whitley

The Multi-Prize Lottery Ticket Hypothesis posits that randomly initialized neural networks contain several subnetworks that achieve comparable accuracy to fully trained models of the same architecture. However, current methods require that the network is sufficiently overparameterized. In this work, we propose a modification to two state-of-the-art algorithms (Edge-Popup and Biprop) that finds high-accuracy subnetworks with no additional storage cost or scaling. The algorithm, Iterative Weight Recycling, identifies subsets of important weights within a randomly initialized network for intra-layer reuse. Empirically we show improvements on smaller network architectures and higher prune rates, finding that model sparsity can be increased through the "recycling" of existing weights. In addition to Iterative Weight Recycling, we complement the Multi-Prize Lottery Ticket Hypothesis with a reciprocal finding: high-accuracy, randomly initialized subnetwork's produce diverse masks, despite being generated with the same hyperparameter's and pruning strategy. We explore the landscapes of these masks, which show high variability.

Tim Whitaker, Darrell Whitley

Rectified Linear Units (ReLU) are the default choice for activation functions in deep neural networks. While they demonstrate excellent empirical performance, ReLU activations can fall victim to the dead neuron problem. In these cases, the weights feeding into a neuron end up being pushed into a state where the neuron outputs zero for all inputs. Consequently, the gradient is also zero for all inputs, which means that the weights which feed into the neuron cannot update. The neuron is not able to recover from direct back propagation and model capacity is reduced as those parameters can no longer be further optimized. Inspired by a neurological process of the same name, we introduce Synaptic Stripping as a means to combat this dead neuron problem. By automatically removing problematic connections during training, we can regenerate dead neurons and significantly improve model capacity and parametric utilization. Synaptic Stripping is easy to implement and results in sparse networks that are more efficient than the dense networks they are derived from. We conduct several ablation studies to investigate these dynamics as a function of network width and depth and we conduct an exploration of Synaptic Stripping with Vision Transformers on a variety of benchmark datasets.

Ofer M. Shir, Jonathan Roslund, Darrell Whitley et al.

The Covariance Matrix Adaptation Evolution Strategy (CMA-ES) has been the most successful Evolution Strategy at exploiting covariance information; it uses a form of Principle Component Analysis which, under certain conditions, is suggested to converge to the correct covariance matrix, formulated as the inverse of the mathematically well-defined Hessian matrix. However, in practice, there exist conditions where CMA-ES converges to the global optimum (accomplishing its primary goal) while it does not learn the true covariance matrix (missing an auxiliary objective), likely due to step-size deficiency. These circumstances can involve high-dimensional landscapes with large condition numbers. This paper introduces a novel technique entitled Forced Optimal Covariance Adaptive Learning (FOCAL), with the explicit goal of determining the Hessian at the global basin of attraction. It begins by introducing theoretical foundations to the inverse relationship between the learned covariance and the Hessian matrices. FOCAL is then introduced and demonstrated to retrieve the Hessian matrix with high fidelity on both model landscapes and experimental Quantum Control systems, which are observed to possess a non-separable, non-quadratic search landscape. The recovered Hessian forms are corroborated by physical knowledge of the systems. This study constitutes an example for Natural Computing successfully serving other branches of natural sciences, and introducing at the same time a powerful generic method for any high-dimensional continuous search seeking landscape information.

Tim Whitaker, Darrell Whitley

Diversity is an important consideration in the construction of robust neural network ensembles. A collection of well trained models will generalize better if they are diverse in the patterns they respond to and the predictions they make. Diversity is especially important for low-cost ensemble methods because members often share network structure in order to avoid training several independent models from scratch. Diversity is traditionally analyzed by measuring differences between the outputs of models. However, this gives little insight into how knowledge representations differ between ensemble members. This paper introduces several interpretability methods that can be used to qualitatively analyze diversity. We demonstrate these techniques by comparing the diversity of feature representations between child networks using two low-cost ensemble algorithms, Snapshot Ensembles and Prune and Tune Ensembles. We use the same pre-trained parent network as a starting point for both methods which allows us to explore how feature representations evolve over time. This approach to diversity analysis can lead to valuable insights and new perspectives for how we measure and promote diversity in ensemble methods.

Swetha Varadarajan, Darrell Whitley

The travelling salesman problem (TSP) is one of the well-studied NP-hard problems in the literature. The state-of-the art inexact TSP solvers are the Lin-Kernighan-Helsgaun (LKH) heuristic and Edge Assembly crossover (EAX). A recent study suggests that EAX with restart mechanisms perform well on a wide range of TSP instances. However, this study is limited to 2,000 city problems. We study for problems ranging from 2,000 to 85,900. We see that the performance of the solver varies with the type of the problem. However, combining these solvers in an ensemble setup, we are able to outperform the individual solver's performance. We see the ensemble setup as an efficient way to make use of the abundance of compute resources. In addition to EAX and LKH, we use several versions of the hybrid of EAX and Mixing Genetic Algorithm (MGA). A hybrid of MGA and EAX is known to solve some hard problems. We see that the ensemble of the hybrid version outperforms the state-of-the-art solvers on problems larger than 10,000 cities.

Jonathan Heins, Lennart Schäpermeier, Pascal Kerschke et al.

Solving the Traveling Salesperson Problem (TSP) remains a persistent challenge, despite its fundamental role in numerous generalized applications in modern contexts. Heuristic solvers address the demand for finding high-quality solutions efficiently. Among these solvers, the Lin-Kernighan-Helsgaun (LKH) heuristic stands out, as it complements the performance of genetic algorithms across a diverse range of problem instances. However, frequent timeouts on challenging instances hinder the practical applicability of the solver. Within this work, we investigate a previously overlooked factor contributing to many timeouts: The use of a fixed candidate set based on a tree structure. Our investigations reveal that candidate sets based on Hamiltonian circuits contain more optimal edges. We thus propose to integrate this promising initialization strategy, in the form of POPMUSIC, within an efficient restart version of LKH. As confirmed by our experimental studies, this refined TSP heuristic is much more efficient - causing fewer timeouts and improving the performance (in terms of penalized average runtime) by an order of magnitude - and thereby challenges the state of the art in TSP solving.

Renato Tinós, Liang Zhao, Francisco Chicano et al.

The NK hybrid genetic algorithm for clustering is proposed in this paper. In order to evaluate the solutions, the hybrid algorithm uses the NK clustering validation criterion 2 (NKCV2). NKCV2 uses information about the disposition of $N$ small groups of objects. Each group is composed of $K+1$ objects of the dataset. Experimental results show that density-based regions can be identified by using NKCV2 with fixed small $K$. In NKCV2, the relationship between decision variables is known, which in turn allows us to apply gray box optimization. Mutation operators, a partition crossover, and a local search strategy are proposed, all using information about the relationship between decision variables. In partition crossover, the evaluation function is decomposed into $q$ independent components; partition crossover then deterministically returns the best among $2^q$ possible offspring with computational complexity $O(N)$. The NK hybrid genetic algorithm allows the detection of clusters with arbitrary shapes and the automatic estimation of the number of clusters. In the experiments, the NK hybrid genetic algorithm produced very good results when compared to another genetic algorithm approach and to state-of-art clustering algorithms.

Jonathan Heins, Darrell Whitley, Pascal Kerschke

The Edge Assembly Crossover (EAX) algorithm is the state-of-the-art heuristic for solving the Traveling Salesperson Problem (TSP). It regularly outperforms other methods, such as the Lin-Kernighan-Helsgaun heuristic (LKH), across diverse sets of TSP instances. Essentially, EAX employs a two-stage mechanism that focuses on improving the current solutions, first, at the local and, subsequently, at the global level. Although the second phase of the algorithm has been thoroughly studied, configured, and refined in the past, in particular, its first stage has hardly been examined. In this paper, we thus focus on the first stage of EAX and introduce a novel method that quickly verifies whether the AB-cycles, generated during its internal optimization procedure, yield valid tours -- or whether they need to be repaired. Knowledge of the latter is also particularly relevant before applying other powerful crossover operators such as the Generalized Partition Crossover (GPX). Based on our insights, we propose and evaluate several improved versions of EAX. According to our benchmark study across 10 000 different TSP instances, the most promising of our proposed EAX variants demonstrates improved computational efficiency and solution quality on previously rather difficult instances compared to the current state-of-the-art EAX algorithm.

Tim Whitaker, Darrell Whitley

Pruning methods have recently grown in popularity as an effective way to reduce the size and computational complexity of deep neural networks. Large numbers of parameters can be removed from trained models with little discernible loss in accuracy after a small number of continued training epochs. However, pruning too many parameters at once often causes an initial steep drop in accuracy which can undermine convergence quality. Iterative pruning approaches mitigate this by gradually removing a small number of parameters over multiple epochs. However, this can still lead to subnetworks that overfit local regions of the loss landscape. We introduce a novel and effective approach to tuning subnetworks through a regularization technique we call Stochastic Subnetwork Annealing. Instead of removing parameters in a discrete manner, we instead represent subnetworks with stochastic masks where each parameter has a probabilistic chance of being included or excluded on any given forward pass. We anneal these probabilities over time such that subnetwork structure slowly evolves as mask values become more deterministic, allowing for a smoother and more robust optimization of subnetworks at high levels of sparsity.

Tim Whitaker, Darrell Whitley

Ensemble Learning is an effective method for improving generalization in machine learning. However, as state-of-the-art neural networks grow larger, the computational cost associated with training several independent networks becomes expensive. We introduce a fast, low-cost method for creating diverse ensembles of neural networks without needing to train multiple models from scratch. We do this by first training a single parent network. We then create child networks by cloning the parent and dramatically pruning the parameters of each child to create an ensemble of members with unique and diverse topologies. We then briefly train each child network for a small number of epochs, which now converge significantly faster when compared to training from scratch. We explore various ways to maximize diversity in the child networks, including the use of anti-random pruning and one-cycle tuning. This diversity enables "Prune and Tune" ensembles to achieve results that are competitive with traditional ensembles at a fraction of the training cost. We benchmark our approach against state of the art low-cost ensemble methods and display marked improvement in both accuracy and uncertainty estimation on CIFAR-10 and CIFAR-100.

Francisco Chicano, Darrell Whitley, Renato Tinos

Local search algorithms and iterated local search algorithms are a basic technique. Local search can be a stand along search methods, but it can also be hybridized with evolutionary algorithms. Recently, it has been shown that it is possible to identify improving moves in Hamming neighborhoods for k-bounded pseudo-Boolean optimization problems in constant time. This means that local search does not need to enumerate neighborhoods to find improving moves. It also means that evolutionary algorithms do not need to use random mutation as a operator, except perhaps as a way to escape local optima. In this paper, we show how improving moves can be identified in constant time for multiobjective problems that are expressed as k-bounded pseudo-Boolean functions. In particular, multiobjective forms of NK Landscapes and Mk Landscapes are considered.

Darrell Whitley, Renato Tinós, Francisco Chicano

This paper introduces the NK Echo State Network. The problem of learning in the NK Echo State Network is reduced to the problem of optimizing a special form of a Spin Glass Problem known as an NK Landscape. No weight adjustment is used; all learning is accomplished by spinning up (turning on) or spinning down (turning off) neurons in order to find a combination of neurons that work together to achieve the desired computation. For special types of NK Landscapes, an exact global solution can be obtained in polynomial time using dynamic programming. The NK Echo State Network is applied to a reinforcement learning problem requiring a recurrent network: balancing two poles on a cart given no velocity information. Empirical results shows that the NK Echo State Network learns very rapidly and yields very good generalization.