Mark Daley

Jacob Morra, Andrew Flynn, Andreas Amann et al.

Multifunctionality describes the capacity for a neural network to perform multiple mutually exclusive tasks without altering its network connections; and is an emerging area of interest in the reservoir computing machine learning paradigm. Multifunctionality has been observed in the brains of humans and other animals: particularly, in the lateral horn of the fruit fly. In this work, we transplant the connectome of the fruit fly lateral horn to a reservoir computer (RC), and investigate the extent to which this 'fruit fly RC' (FFRC) exhibits multifunctionality using the 'seeing double' problem as a benchmark test. We furthermore explore the dynamics of how this FFRC achieves multifunctionality while varying the network's spectral radius. Compared to the widely-used Erdös-Renyi Reservoir Computer (ERRC), we report that the FFRC exhibits a greater capacity for multifunctionality; is multifunctional across a broader hyperparameter range; and solves the seeing double problem far beyond the previously observed spectral radius limit, wherein the ERRC's dynamics become chaotic.

Jacob Morra, Mark Daley

We report an improvement to the conventional Echo State Network (ESN) across three benchmark chaotic time-series prediction tasks using fruit fly connectome data alone. We also investigate the impact of key connectome-derived structural features on prediction performance -- uniquely bridging neurobiological structure and machine learning function; and find that both increasing the global average clustering coefficient and modifying the position of weights -- by permuting their synapse-synapse partners -- can lead to increased model variance and (in some cases) degraded performance. In all we consider four topological point modifications to a connectome-derived ESN reservoir (null model): namely, we alter the network sparsity, re-draw nonzero weights from a uniform distribution, permute nonzero weight positions, and increase the network global average clustering coefficient. We compare the four resulting ESN model classes -- and the null model -- with a conventional ESN by conducting time-series prediction experiments on size-variants of the Mackey-Glass 17 (MG-17), Lorenz, and Rossler chaotic time series; denoting each model's performance and variance across train-validate trials.

Jacob Morra, Mark Daley

Can connectome-derived constraints inform computation? In this paper we investigate the contribution of a fruit fly connectome's topology on the performance of an Echo State Network (ESN) -- a subset of Reservoir Computing which is state of the art in chaotic time series prediction. Specifically, we replace the reservoir layer of a classical ESN -- normally a fixed, random graph represented as a 2-d matrix -- with a particular (female) fruit fly connectome-derived connectivity matrix. We refer to this experimental class of models (with connectome-derived reservoirs) as "Fruit Fly ESNs" (FFESNs). We train and validate the FFESN on a chaotic time series prediction task; here we consider four sets of trials with different training input sizes (small, large) and train-validate splits (two variants). We compare the validation performance (Mean-Squared Error) of all of the best FFESN models to a class of control model ESNs (simply referred to as "ESNs"). Overall, for all four sets of trials we find that the FFESN either significantly outperforms (and has lower variance than) the ESN; or simply has lower variance than the ESN.

Ethan C. Jackson, Mark Daley

Reinforcement learning (RL) problems often feature deceptive local optima, and learning methods that optimize purely for reward signal often fail to learn strategies for overcoming them. Deep neuroevolution and novelty search have been proposed as effective alternatives to gradient-based methods for learning RL policies directly from pixels. In this paper, we introduce and evaluate the use of novelty search over agent action sequences by string edit metric distance as a means for promoting innovation. We also introduce a method for stagnation detection and population resampling inspired by recent developments in the RL community that uses the same mechanisms as novelty search to promote and develop innovative policies. Our methods extend a state-of-the-art method for deep neuroevolution using a simple-yet-effective genetic algorithm (GA) designed to efficiently learn deep RL policy network weights. Experiments using four games from the Atari 2600 benchmark were conducted. Results provide further evidence that GAs are competitive with gradient-based algorithms for deep RL. Results also demonstrate that novelty search over action sequences is an effective source of selection pressure that can be integrated into existing evolutionary algorithms for deep RL.

Ethan C. Jackson, James Alexander Hughes, Mark Daley

In contrast to conventional, univariate analysis, various types of multivariate analysis have been applied to functional magnetic resonance imaging (fMRI) data. In this paper, we compare two contemporary approaches for multivariate regression on task-based fMRI data: linear regression with ridge regularization and non-linear symbolic regression using genetic programming. The data for this project is representative of a contemporary fMRI experimental design for visual stimuli. Linear and non-linear models were generated for 10 subjects, with another 4 withheld for validation. Model quality is evaluated by comparing $R$ scores (Pearson product-moment correlation) in various contexts, including single run self-fit, within-subject generalization, and between-subject generalization. Propensity for modelling strategies to overfit is estimated using a separate resting state scan. Results suggest that neither method is objectively or inherently better than the other.