Kernel density compression for real-time Bayesian encoding/decoding of unsorted hippocampal spikes

Alpha oscillations encode Bayesian belief updating underlying attentional allocation in dynamic environments

In a dynamic environment, expectations of the future constantly change based on updated evidence and affect the dynamic allocation of attention. To further investigate the neural mechanisms underlying attentional expectancies, we employed a modified Central Cue Posner Paradigm in which the probability of cues being valid (that is, accurately indicated the upcoming target location) was manipulated. Attentional deployment to the cued location (α), which was governed by precision of predictions on previous trials, was estimated using a hierarchical Bayesian model and was included as a regressor in the analyses of electrophysiological (EEG) data. Our results revealed that before the target appeared, alpha oscillations (8∼13 Hz) for high-predictability cues (88 % valid) were significantly predicted by precision-dependent attention (α). This relationship was not observed under low-predictability conditions (69 % and 50 % valid cues). After the target appeared, precision-dependent attention (α) correlated with alpha band oscillations only in the valid cue condition and not in the invalid condition. Further analysis under conditions of significant attentional modulation by precision suggested a separate effect of cue orientation. These results provide new insights on how trial-by-trial Bayesian belief updating relates to alpha band encoding of environmentally-sensitive allocation of visual spatial attention.

Exploration vs Convergence Speed in Adaptive-Bias Enhanced Sampling

In adaptive-bias enhanced sampling methods, a bias potential is added to the system to drive transitions between metastable states. The bias potential is a function of a few collective variables and is gradually modified according to the underlying free energy surface. We show that when the collective variables are suboptimal, there is an exploration–convergence tradeoff, and one must choose between a quickly converging bias that will lead to fewer transitions or a slower to converge bias that can explore the phase space more efficiently but might require a much longer time to produce an accurate free energy estimate. The recently proposed on-the-fly probability enhanced sampling (OPES) method focuses on fast convergence, but there are cases where fast exploration is preferred instead. For this reason, we introduce a new variant of the OPES method that focuses on quickly escaping metastable states at the expense of convergence speed. We illustrate the benefits of this approach in prototypical systems and show that it outperforms the popular metadynamics method.

Single-trial dynamics of hippocampal spatial representations are modulated by reward value

Reward value is known to modulate learning speed in spatial memory tasks, but little is known about its influence on the dynamical changes in hippocampal spatial representations. Here, we monitored the trial-to-trial changes in hippocampal place cell activity during the acquisition of place-reward associations with varying reward size. We show a faster reorganization and stabilization of the hippocampal place map when a goal location is associated with a large reward. The reorganization is driven by both rate changes and the appearance and disappearance of place fields. The occurrence of hippocampal replay activity largely followed the dynamics of changes in spatial representations. Replay patterns became more selectively tuned toward behaviorally relevant experiences over the course of learning via the refined contributions of specific cell subpopulations. These results suggest that high reward value enhances memory retention by accelerating the formation and stabilization of the hippocampal cognitive map and selectively enhancing its reactivation during learning.

Assessing Goodness-of-Fit in Marked Point Process Models of Neural Population Coding via Time and Rate Rescaling

Marked point process models have recently been used to capture the coding properties of neural populations from multiunit electrophysiological recordings without spike sorting. These clusterless models have been shown in some instances to better describe the firing properties of neural populations than collections of receptive field models for sorted neurons and to lead to better decoding results. To assess their quality, we previously proposed a goodness-of-fit technique for marked point process models based on time rescaling, which for a correct model produces a set of uniform samples over a random region of space. However, assessing uniformity over such a region can be challenging, especially in high dimensions. Here, we propose a set of new transformations in both time and the space of spike waveform features, which generate events that are uniformly distributed in the new mark and time spaces. These transformations are scalable to multidimensional mark spaces and provide uniformly distributed samples in hypercubes, which are well suited for uniformity tests. We discuss the properties of these transformations and demonstrate aspects of model fit captured by each transformation. We also compare multiple uniformity tests to determine their power to identify lack-of-fit in the rescaled data. We demonstrate an application of these transformations and uniformity tests in a simulation study. Proofs for each transformation are provided in the appendix.

Rethinking Metadynamics: From Bias Potentials to Probability Distributions

Metadynamics is an enhanced sampling method of great popularity, based on the on-the-fly construction of a bias potential that is a function of a selected number of collective variables. We propose here a change in perspective that shifts the focus from the bias to the probability distribution reconstruction while retaining some of the key characteristics of metadynamics, such as flexible on-the-fly adjustments to the free energy estimate. The result is an enhanced sampling method that presents a drastic improvement in convergence speed, especially when dealing with suboptimal and/or multidimensional sets of collective variables. The method is especially robust and easy to use and in fact requires only a few simple parameters to be set, and it has a straightforward reweighting scheme to recover the statistics of the unbiased ensemble. Furthermore, it gives more control of the desired exploration of the phase space since the deposited bias is not allowed to grow indefinitely and it does not push the simulation to uninteresting high free energy regions. We demonstrate the performance of the method in a number of representative examples.

Real-Time Readout of Large-Scale Unsorted Neural Ensemble Place Codes

Uncovering spatial representations from large-scale ensemble spike activity in specific brain circuits provides valuable feedback in closed-loop experiments. We develop a graphics processing unit (GPU)-powered population-decoding system for ultrafast reconstruction of spatial positions from rodents' unsorted spatiotemporal spiking patterns, during run behavior or sleep. In comparison with an optimized quad-core central processing unit (CPU) implementation, our approach achieves an ∼20- to 50-fold increase in speed in eight tested rat hippocampal, cortical, and thalamic ensemble recordings, with real-time decoding speed (approximately fraction of a millisecond per spike) and scalability up to thousands of channels. By accommodating parallel shuffling in real time (computation time <15 ms), our approach enables assessment of the statistical significance of online-decoded "memory replay" candidates during quiet wakefulness or sleep. This open-source software toolkit supports the decoding of spatial correlates or content-triggered experimental manipulation in closed-loop neuroscience experiments.

Real-time classification of experience-related ensemble spiking patterns for closed-loop applications

Communication in neural circuits across the cortex is thought to be mediated by spontaneous temporally organized patterns of population activity lasting ~50 –200 ms. Closed-loop manipulations have the unique power to reveal direct and causal links between such patterns and their contribution to cognition. Current brain–computer interfaces, however, are not designed to interpret multi-neuronal spiking patterns at the millisecond timescale. To bridge this gap, we developed a system for classifying ensemble patterns in a closed-loop setting and demonstrated its application in the online identification of hippocampal neuronal replay sequences in the rat. Our system decodes multi-neuronal patterns at 10 ms resolution, identifies within 50 ms experience-related patterns with over 70% sensitivity and specificity, and classifies their content with 95% accuracy. This technology scales to high-count electrode arrays and will help to shed new light on the contribution of internally generated neural activity to coordinated neural assembly interactions and cognition.

A common goodness-of-fit framework for neural population models using marked point process time-rescaling

A critical component of any statistical modeling procedure is the ability to assess the goodness-of-fit between a model and observed data. For spike train models of individual neurons, many goodness-of-fit measures rely on the time-rescaling theorem and assess model quality using rescaled spike times. Recently, there has been increasing interest in statistical models that describe the simultaneous spiking activity of neuron populations, either in a single brain region or across brain regions. Classically, such models have used spike sorted data to describe relationships between the identified neurons, but more recently clusterless modeling methods have been used to describe population activity using a single model. Here we develop a generalization of the time-rescaling theorem that enables comprehensive goodness-of-fit analysis for either of these classes of population models. We use the theory of marked point processes to model population spiking activity, and show that under the correct model, each spike can be rescaled individually to generate a uniformly distributed set of events in time and the space of spike marks. After rescaling, multiple well-established goodness-of-fit procedures and statistical tests are available. We demonstrate the application of these methods both to simulated data and real population spiking in rat hippocampus. We have made the MATLAB and Python code used for the analyses in this paper publicly available through our Github repository at https://github.com/Eden-Kramer-Lab/popTRT .

Recurrent network model for learning goal-directed sequences through reverse replay

Reverse replay of hippocampal place cells occurs frequently at rewarded locations, suggesting its contribution to goal-directed path learning. Symmetric spike-timing dependent plasticity (STDP) in CA3 likely potentiates recurrent synapses for both forward (start to goal) and reverse (goal to start) replays during sequential activation of place cells. However, how reverse replay selectively strengthens forward synaptic pathway is unclear. Here, we show computationally that firing sequences bias synaptic transmissions to the opposite direction of propagation under symmetric STDP in the co-presence of short-term synaptic depression or afterdepolarization. We demonstrate that significant biases are created in biologically realistic simulation settings, and this bias enables reverse replay to enhance goal-directed spatial memory on a W-maze. Further, we show that essentially the same mechanism works in a two-dimensional open field. Our model for the first time provides the mechanistic account for the way reverse replay contributes to hippocampal sequence learning for reward-seeking spatial navigation.

Falcon: a highly flexible open-source software for closed-loop neuroscience

OBJECTIVE

Closed-loop experiments provide unique insights into brain dynamics and function. To facilitate a wide range of closed-loop experiments, we created an open-source software platform that enables high-performance real-time processing of streaming experimental data.

APPROACH

We wrote Falcon, a C++ multi-threaded software in which the user can load and execute an arbitrary processing graph. Each node of a Falcon graph is mapped to a single thread and nodes communicate with each other through thread-safe buffers. The framework allows for easy implementation of new processing nodes and data types. Falcon was tested both on a 32-core and a 4-core workstation. Streaming data was read from either a commercial acquisition system (Neuralynx) or the open-source Open Ephys hardware, while closed-loop TTL pulses were generated with a USB module for digital output. We characterized the round-trip latency of our Falcon-based closed-loop system, as well as the specific latency contribution of the software architecture, by testing processing graphs with up to 32 parallel pipelines and eight serial stages. We finally deployed Falcon in a task of real-time detection of population bursts recorded live from the hippocampus of a freely moving rat.

MAIN RESULTS

On Neuralynx hardware, round-trip latency was well below 1 ms and stable for at least 1 h, while on Open Ephys hardware latencies were below 15 ms. The latency contribution of the software was below 0.5 ms. Round-trip and software latencies were similar on both 32- and 4-core workstations. Falcon was used successfully to detect population bursts online with ~40 ms average latency.

SIGNIFICANCE

Falcon is a novel open-source software for closed-loop neuroscience. It has sub-millisecond intrinsic latency and gives the experimenter direct control of CPU resources. We envisage Falcon to be a useful tool to the neuroscientific community for implementing a wide variety of closed-loop experiments, including those requiring use of complex data structures and real-time execution of computationally intensive algorithms, such as population neural decoding/encoding from large cell assemblies.