Model for a flexible motor memory based on a self-active recurrent neural network

Boström, Kim Joris; Wagner, Heiko; Prieske, M.; Lussanet, Marc H. E. de

doi:10.1016/j.humov.2013.07.003

Cited by 9 publications

(7 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In fsRNNs, feedback from the read-out unit to the neurons in the RNN should be comparable in strength to the recurrent inputs to those neurons in order to suppress chaos and allow for successful training 9 . In our implementation, the strength of the feedback is governed by the parameter g FB , which is typically set to be around 1 in applications of FORCE training RNNs 9 , 37 – 39 . We varied this number from 0 to 5 and trained 20 networks for each value.…”

Section: Resultsmentioning

confidence: 99%

Flexibility in motor timing constrains the topology and dynamics of pattern generator circuits

2018

View full text Add to dashboard Cite

Temporally precise movement patterns underlie many motor skills and innate actions, yet the flexibility with which the timing of such stereotyped behaviors can be modified is poorly understood. To probe this, we induce adaptive changes to the temporal structure of birdsong. We find that the duration of specific song segments can be modified without affecting the timing in other parts of the song. We derive formal prescriptions for how neural networks can implement such flexible motor timing. We find that randomly connected recurrent networks, a common approximation for how neocortex is wired, do not generally conform to these, though certain implementations can approximate them. We show that feedforward networks, by virtue of their one-to-one mapping between network activity and time, are better suited. Our study provides general prescriptions for pattern generator networks that implement flexible motor timing, an important aspect of many motor skills, including birdsong and human speech.

show abstract

Section: Resultsmentioning

confidence: 99%

Flexibility in motor timing constrains the topology and dynamics of pattern generator circuits

2018

View full text Add to dashboard Cite

show abstract

“…For example, the reservoir activity can be read out to reproduce the trajectories of markers placed on the human body during walking and running (Sussillo & Abbott, 2009). Reservoir computing has been used to model how biological neural networks achieve a variety of tasks (Hinaut & Dominey, 2013; Boström, 2013; Wyffels & Schrauwen, 2009). It provides an interesting model as it captures three important properties of biological neural networks: (1) they are recurrent (i.e.…”

Section: Introductionmentioning

confidence: 99%

“…Within the reservoir computing framework, EI balance has been modelled with varying degrees of biological realism. Echo state networks typically do not have distinct excitatory and inhibitory neurons: the outgoing connection weights of individual neurons are most often drawn from a random distribution centered around zero (Sussillo & Abbott, 2009; Boström, 2013; Jaeger, 2010). As a result, any given neuron exerts both excitatory and inhibitory influences.…”

Section: Introductionmentioning

confidence: 99%

Motor pattern generation is robust to neural network anatomical imbalance favoring inhibition but not excitation

Graaf

Mochizuki

Thies

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Animals display rich and coordinated motor patterns during walking and running. Previous modelling as well as experimental results suggest that the balance between excitation and inhibition in neural networks may be critical for generating such structured motor patterns. However, biological neural networks have an anatomical imbalance between excitatory and inhibitory neural populations. We explore the influence of such an anatomical imbalance on the ability of a reservoir computing artificial neural network to learn human locomotor patterns for slow walking, fast walking and running. We varied the numbers of neurons, connections percentages and connection strengths of excitatory and inhibitory populations. We showed that performance depended on the network anatomy. First, it deteriorated when the total number of neurons was too small or the total connection strength was too large. Second, performance was critically dependent on the balance between excitation and inhibition. Imbalance towards excitation caused a reduction in the richness of internal network dynamics, leading to a stereotypical motor output and poor overall performance. In contrast, rich internal dynamics and good overall performance were found when the network anatomy was either balanced or imbalanced towards inhibition. This suggests that motor pattern generation may be robust to increased inhibition but not increased excitation in neural networks.1Author summaryHow does the anatomy of the nervous system allow the generation of the complex motor patterns observed during the movements of humans and other animals? We explore this question in a model of the spinal cord in which we vary the neural anatomy. We find that movement generation requires the neural network to have rich internal dynamics. Such rich internal dynamics emerge from the interaction between the excitatory and inhibitory neurons in the network. Strong inhibition causes fluctuations in the neural activity which allow rich motor patterns to be produced. However, strong excitation quenches these fluctuations and causes a reduction in the variability of motor patterns. When both excitation and inhibition are strong, the neural activity becomes chaotic, and dysfunctional, highly variable motor patterns are produced. We therefore predict that diseases of the nervous system which affect inhibitory and excitatory neurons differently will have a different signature in terms of motor patterns. Diseases causing increased excitation in neural circuits should lead to stereotypical motor behaviors, whereas diseases causing increased excitation and inhibition should lead to unstable motor patterns.

show abstract

“…Another popular method of processing EMG signals is features extractions [18] and machine learning, but this requires tailoring the features [19] based on the problem at hand and the training is not end-to-end. Deep learning [20] methodologies, and non-spiking reservoir computing [21] have also been studied in the past for EMG signal processing, but such architectures are power-hungry and difficult to deploy on edge in real-time; moreover, the performance of deep neural networks is limited by the availability of large datasets. While there are studies done in the past to evaluate the performance of plastic spiking reservoirs for processing electroencephalogram (EEG) signals [22], such a study of EMG signal processing is not known to the authors.…”

Section: Introductionmentioning

confidence: 99%

Signals to Spikes for Neuromorphic Regulated Reservoir Computing and EMG Hand Gesture Recognition

Garg,

Balafrej,

Beilliard

et al. 2021

Preprint

View full text Add to dashboard Cite

Surface electromyogram (sEMG) signals result from muscle movement and hence they are an ideal candidate for benchmarking event-driven sensing and computing. We propose a simple yet novel approach for optimizing the spike encoding algorithm's hyperparameters inspired by the readout layer concept in reservoir computing. Using a simple machine learning algorithm after spike encoding, we report performance higher than the state-of-the-art spiking neural networks on two open-source datasets for hand gesture recognition. The spike encoded data is processed through a spiking reservoir with a biologically inspired topology and neuron model. When trained with the unsupervised activity regulation CRITICAL algorithm to operate at the edge of chaos, the reservoir yields better performance than state-of-the-art convolutional neural networks. The reservoir performance with regulated activity was found to be 89.72% for the Roshambo EMG dataset and 70.6% for the EMG subset of sensor fusion dataset. Therefore, the biologicallyinspired computing paradigm, which is known for being power efficient, also proves to have a great potential when compared with conventional AI algorithms. CCS CONCEPTS• Computing methodologies → Bio-inspired approaches.

show abstract

Model for a flexible motor memory based on a self-active recurrent neural network

Cited by 9 publications

References 51 publications

Flexibility in motor timing constrains the topology and dynamics of pattern generator circuits

Flexibility in motor timing constrains the topology and dynamics of pattern generator circuits

Motor pattern generation is robust to neural network anatomical imbalance favoring inhibition but not excitation

Signals to Spikes for Neuromorphic Regulated Reservoir Computing and EMG Hand Gesture Recognition

Contact Info

Product

Resources

About