Probing latent brain dynamics in Alzheimer's disease via recurrent neural network

Chaotic recurrent neural networks for brain modelling: A review

Even in the absence of external stimuli, the brain is spontaneously active. Indeed, most cortical activity is internally generated by recurrence. Both theoretical and experimental studies suggest that chaotic dynamics characterize this spontaneous activity. While the precise function of brain chaotic activity is still puzzling, we know that chaos confers many advantages. From a computational perspective, chaos enhances the complexity of network dynamics. From a behavioural point of view, chaotic activity could generate the variability required for exploration. Furthermore, information storage and transfer are maximized at the critical border between order and chaos. Despite these benefits, many computational brain models avoid incorporating spontaneous chaotic activity due to the challenges it poses for learning algorithms. In recent years, however, multiple approaches have been proposed to overcome this limitation. As a result, many different algorithms have been developed, initially within the reservoir computing paradigm. Over time, the field has evolved to increase the biological plausibility and performance of the algorithms, sometimes going beyond the reservoir computing framework. In this review article, we examine the computational benefits of chaos and the unique properties of chaotic recurrent neural networks, with a particular focus on those typically utilized in reservoir computing. We also provide a detailed analysis of the algorithms designed to train chaotic RNNs, tracing their historical evolution and highlighting key milestones in their development. Finally, we explore the applications and limitations of chaotic RNNs for brain modelling, consider their potential broader impacts beyond neuroscience, and outline promising directions for future research.

Dynamics of interaction between IH and IKLT currents to mediate double resonances of medial superior olive neurons related to sound localization

Neurons in the medial superior olive (MSO) exhibit high frequency responses such as subthreshold resonance, which is helpful to sensitively detect a small difference in the arrival time of sounds between two ears for precise sound localization. Recently, except for the high frequency depolarization resonance mediated by a low threshold potassium (IKLT) current, a low frequency hyperpolarization resonance mediated by a hyperpolarization-activated cation (IH) current is observed in experiments on the MSO neurons, forming double resonances. The complex dynamics underlying double resonances are studied in an MSO neuron model in the present paper. Firstly, double resonances similar to the experimental observations are simulated as the resting membrane potential is between half-activation voltages of IH and IKLT currents, and stimulation current (IZAP) with large amplitude and exponentially increasing frequency is applied. Secondly, multiple effective factors to modulate double resonances are obtained. Especially, the decrease of time constant of IKLT current and increase of conductance of IH and IKLT currents can enhance the depolarization resonance frequency for precise sound localization. Last, different frequency responses of slow IH and fast IKLT currents in formation of the resonances are acquired. A middle phase difference between IZAP and IKLT currents appears at a high frequency, and the interaction between the positive part of IZAP and the negative IKLT current forms the depolarization resonance. Interaction between the negative part of IZAP and positive IH current with a middle phase difference results in hyperpolarization resonance at a low frequency. Furthermore, the phase difference between IZAP and resonance current can well explain the increase of depolarization resonance frequency modulated by the increase of conductance of IH or IKLT currents. The results present the dynamical and biophysical mechanisms for the double resonances mediated by two currents in the MSO neurons, which is helpful to enhance the depolarization resonance frequency for precise sound localization.