Dan F. M. Goodman

Factorization and spatial encodings: a hypothesis about the foundations of the genomic code

The paper proposes that spatial encoding and factorisation mechanisms enable efficient genomic encoding of neural network architecture and connectivity.

Partial recurrence enables robust and efficient computation

This paper introduces partially recurrent neural networks (pRNNs) and shows they can match or exceed fully recurrent networks in performance while using fewer parameters.

Neuromodulation enhances the capability and efficiency of spiking neural networks

This study shows neuromodulation improves spiking neural network performance and efficiency in sensory tasks with minimal resource use.

Spiking neural network models of sound localisation via a massively collaborative process

The paper describes a massively collaborative project using spiking neural networks to study sound localization, involving 31 researchers across multiple countries.

Fusing multisensory signals across channels and time

The paper introduces a novel multisensory task to evaluate how well algorithms can integrate signals across sensory channels and time, showing that simple temporal integration can match complex recurrent models.

Nonlinear fusion is optimal for a wide class of multisensory tasks

The paper demonstrates that nonlinear fusion outperforms linear fusion in multisensory integration tasks when sensory signals are statistically dependent, and shows that spiking neural networks can approximate this nonlinear algorithm.

Neural heterogeneity promotes robust learning

This study shows that neural heterogeneity improves task performance and learning stability in spiking neural networks, matching experimental parameter distributions.

The Psychometrics of Automatic Speech Recognition

This study evaluates how well deep learning-based automatic speech recognition systems match human performance in psychometric experiments, finding that while some systems show agreement, none fully replicate human behavior.

Canonical brain computations account for perceived sound source location

The paper tests neural mechanisms for sound localization in reverberant environments and finds that common brain circuits can reproduce localization preferences without specialized mechanisms.

Short-term effects of sound localization training in virtual reality

This study evaluates three virtual sound localization training paradigms, finding that active listening significantly improves localization accuracy, with effects generalizing to new HRTFs.

Brian 2: an intuitive and efficient neural simulator

Brian 2 is a neural simulator that simplifies model definition and improves performance through runtime code generation, demonstrated with various examples.

Brian2GeNN: a system for accelerating alarge variety of spiking neural networkswith graphics hardware

Brian2GeNN accelerates spiking neural network simulations by translating Brian scripts into GPU-accelerated C++ code via GeNN, improving performance by tens to hundreds of times.

Modeling neuron-glia interactions with the Brian 2 simulator

The paper demonstrates how to model neuron-glia interactions using the Brian 2 simulator, highlighting its flexibility for simulating glial physiology and astrocyte influence on neural activity.

Modelling firing regularity in the ventral cochlear nucleus: mechanisms, and effects of stimulus level and synaptopathy

This paper develops a low-parameter model to study firing regularity in chopper cells of the ventral cochlear nucleus, showing multiple mechanisms can explain their behavior and predicting effects of synaptopathy.