Where is all the nonlinearity: flexible nonlinear modeling of behaviorally relevant neural dynamics using recurrent neural networks

Understanding the dynamical transformation of neural activity to behavior requires modeling this transformation while both dissecting its potential nonlinearities and dissociating and preserving its nonlinear behaviorally relevant neural dynamics, which remain unaddressed. We present RNN PSID, a nonlinear dynamic modeling method that enables flexible dissection of nonlinearities, dissociation and preferential learning of neural dynamics relevant to specific behaviors, and causal decoding. We first validate RNN PSID in simulations and then use it to investigate nonlinearities in monkey spiking and LFP activity across four tasks and different brain regions. Nonlinear RNN PSID successfully dissociated and preserved nonlinear behaviorally relevant dynamics, thus outperforming linear and non-preferential nonlinear learning methods in behavior decoding while reaching similar neural prediction. Strikingly, dissecting the nonlinearities with RNN PSID revealed that consistently across all tasks, summarizing the nonlinearity only in the mapping from the latent dynamics to behavior was largely sufficient for predicting behavior and neural activity. RNN PSID provides a novel tool to reveal new characteristics of nonlinear neural dynamics underlying behavior.

Multibranch Formal Neuron: An Internally Nonlinear Learning Unit

The transformation of synaptic input into action potential in nerve cells is strongly influenced by the morphology of the dendritic arbor as well as the synaptic efficacy map. The multiplicity of dendritic branches strikingly enables a single cell to act as a highly nonlinear processing element. Studies have also found functional synaptic clustering whereby synapses that encode a common sensory feature are spatially clustered together on the branches. Motivated by these findings, here we introduce a multibranch formal model of the neuron that can integrate synaptic inputs nonlinearly through collective action of its dendritic branches and yields synaptic clustering. An analysis in support of its use as a computational building block is offered. Also offered is an accompanying gradient descent–based learning algorithm. The model unit spans a wide spectrum of nonlinearities, including the parity problem, and can outperform the multilayer perceptron in generalizing to unseen data. The occurrence of synaptic clustering boosts the generalization efficiency of the unit, which may also be the answer for the puzzling ubiquity of synaptic clustering in the real neurons. Our theoretical analysis is backed up by simulations. The study could pave the way to new artificial neural networks.

Consumer Imperfect Information and Endogenous Price Rigidity

This paper studies the propagation of monetary shocks in an economy featuring a strategic microfoundation for price rigidities. Following an aggregate shock to money, most consumers are initially uninformed. The market for goods is decentralized. Firms are better off delaying the adjustment of prices until enough consumers learn. At the same time, consumers learn from firms that have adjusted prices. The implied endogenous information diffusion follows a Bernoulli differential equation, implying a nonlinear path of learning. Nonlinear learning implies hump-shaped dynamics of output and inflation. A quantitative exercise suggests that these dynamics can be sizable and persistent. (JEL D11, D21, D40, D82, E23, E31)

Queue-based features for dynamic waiting time prediction in emergency department

Purpose The purpose of this study is twofold: exploring new queue-based variables enabled by process mining and evaluating their impact on the accuracy of waiting time prediction. Such queue-based predictors that capture the current state of the emergency department (ED) may lead to a significant improvement in the accuracy of the prediction models. Design/methodology/approach Alongside the traditional variables influencing ED waiting time, the authors developed new queue-based predictors exploiting process mining. Process mining techniques allowed the authors to discover the actual patient-flow and derive information about the crowding level of the activities. The proposed predictors were evaluated using linear and nonlinear learning techniques. The authors used real data from an ED. Findings As expected, the main results show that integrating the set of predictors with queue-based variables significantly improves the accuracy of waiting time prediction. Specifically, mean square error values were reduced by about 22 and 23 per cent by applying linear and nonlinear learning techniques, respectively. Practical implications Accurate estimates of waiting time can enable the ED systems to prevent overcrowding e.g. improving the routing of patients in EDs and managing more efficiently the resources. Providing accurate waiting time information also can lead to decreased patients’ dissatisfaction and elopement. Originality/value The novelty of the study relies on the attempt to derive queue-based variables reporting the crowding level of the activities within the ED through process mining techniques. Such information is often unavailable or particularly difficult to extract automatically, due to the characteristics of ED processes.