Search

We propose that the neocortex implements active predictive coding (APC), a form of predictive coding that incorporates hierarchical dynamics and actions. In this model, each neocortical area estimates both sensory states and actions, and the cortex as whole learns to predict the sensory consequences of actions at multiple hierarchical levels. “Higher” cortical areas maintain more abstract representations at larger spatiotemporal scales compared to “lower” areas. Feedback from higher areas modulate the dynamics of both state and action networks in lower areas. This allows the cortical network to model the complex dynamics and physics of the world in terms of simpler compositional elements (state transition functions). Simultaneously, current higher level goals invoke sequences of lower level sub-goals and actions, allowing the network to solve complex planning problems by composing simpler solutions. Planning (“system 2” thinking) in turns allows the network to learn, over time, perception-to-action mappings (policies; “system 1” thinking) at multiple abstraction levels. We provide examples from simulations illustrating how the same APC architecture can solve problems that, at first blush, seem very different from each other: (1) how do we recognize an object and its parts using eye movements? (2) why does perception seem stable despite eye movements? (3) how do we learn compositional representations, e.g., part-whole hierarchies, and nested reference frames for equivariant vision? (4) how do we model the “physics” of a complex environment by decomposing it into simpler components? (5) how do we plan actions in a complex domain to achieve a goal by composing sequences of sub-goals and simpler actions? and (6) how do we form episodic memories of sensory-motor experiences? We propose a mapping of the APC network to the laminar architecture of the cortex and suggest possible roles for cortico-cortical, cortico-thalamic, cortico-hippocampal and cortico-subcortical pathways.

Objective: A major challenge in designing closed-loop brain-computer interfaces is finding optimal stimulation patterns as a function of ongoing neural activity for different subjects and objectives. Approach: To achieve goal-directed closed-loop neurostimulation, we propose "neural co-processors" which use artificial neural networks and deep learning to learn optimal closed-loop stimulation policies, shaping neural activity and bridging injured neural circuits for targeted repair and rehabilitation. The co-processor adapts the stimulation policy as the biological circuit itself adapts to the stimulation, achieving a form of brain-device co-adaptation. Here we use simulations to lay the groundwork for future in vivo tests of neural co-processors. We leverage a cortical model of grasping, to which we applied various forms of simulated lesions, allowing us to develop the critical learning algorithms and study adaptations to non-stationarity. Main results: Our simulations show the ability of a neural co-processor to learn a stimulation policy using a supervised learning approach, and to adapt that policy as the underlying brain and sensors change. Our co-processor successfully co-adapted with the simulated brain to accomplish the reach-and-grasp task after a variety of lesions were applied, achieving recovery towards healthy function. Significance: Our results provide the first proof-of-concept demonstration of a co-processor for adaptive activity-dependent closed-loop neurostimulation, optimizing for a rehabilitation goal. While a gap remains between simulations and applications, our results provide insights on how co-processors may be developed for learning complex adaptive stimulation policies for a variety of neural rehabilitation and neuroprosthetic applications., 45 pages, 19 figures. Submitted the IOP Journal of Neural Engineering

We introduce dynamic predictive coding, a hierarchical model of spatiotemporal prediction and sequence learning in the cortex. The model assumes that higher cortical levels modulate the temporal dynamics of lower levels, correcting their predictions of dynamics using prediction errors. As a result, lower levels form representations that encode sequences at shorter timescales (e.g., a single step) while higher levels form representations that encode sequences at longer timescales (e.g., an entire sequence). We tested this model using a two-level neural network, where the top-down modulation is implemented as a low-dimensional mixture of possible temporal dynamics. When trained on natural videos, the lower-level model neurons developed space-time receptive fields similar to those of simple cells in the primary visual cortex while the higher-level responses spanned longer timescales, mimicking temporal response hierarchies in the cortex. Additionally, the network’s hierarchical sequence representation exhibited both predictive and postdictive effects resembling those observed in visual motion processing in humans (e.g., the flash-lag illusion). When coupled with an associative memory mimicking the role of the hippocampus, the model allowed episodic memories to be stored and retrieved, supporting cue-triggered recall of an input sequence similar to activity recall in the visual cortex. Taken together, our results suggest that cortical processing and learning of sequences can be interpreted as dynamic predictive coding based on a hierarchical spatiotemporal generative model of the visual world.

Different sleep states have been shown to be vital for a variety of brain function, including learning, memory, and skill consolidation. However, our understanding of neural dynamics during sleep and the role of prominent LFP frequency bands remain incomplete. To elucidate such dynamics and changes between different behavioral states we collected multichannel LFP and spike data in primary motor cortex of unconstrained macaques for up to 24 hours using the Neurochip3. Each 8 second bin of time was classified into awake and moving (Move), awake and at rest (Rest), REM sleep (REM), or non-REM sleep (NREM) by using dimensionality reduction and clustering on the average spectral density and the acceleration of the head. LFP power showed high delta during NREM, high theta during REM, and high beta when the animal was awake. Cross-frequency phase-amplitude coupling typically showed higher coupling for deeper sleep between all pairs of frequency bands. Two notable exceptions were high delta-high gamma and theta-high gamma coupling during Move, and high theta-beta coupling during REM. Sorted single units showed decreased firing rate with deeper sleep, though with higher “bursty” patterns during NREM compared to other states. Spike-LFP synchrony showed high delta synchrony during Move, and higher coupling with all other frequency bands with deeper sleep.These results altogether are consistent with previous findings showing reactivation of cortical circuitry during sleep, which may be moderated by delta band LFP.

Tracking a turbulent plume to locate its source under variable wind and plume statistics is a complex task; flying insects routinely accomplish such tracking, often over long distances, in pursuit of food or mates. Several aspects of this remarkable behavior and its underlying neural circuitry have been studied experimentally. Here, we take a complementary in silico approach to develop an integrated understanding of behavior and neural computations. Specifically, we train artificial recurrent neural network (RNN) agents using deep reinforcement learning (DRL) to locate the source of simulated turbulent plumes. Interestingly, the agents' emergent behaviors resemble those of flying insects, and the RNNs learn to compute task-relevant variables with distinct dynamic structures in population activity. Our analyses put forward a testable behavioral hypothesis for tracking plumes in changing wind direction, and we provide key intuitions for memory requirements and neural dynamics in turbulent plume tracking.

Intracortical microstimulation (ICMS) is commonly used in many experimental and clinical paradigms; however, its effects on the activation of neurons are still not completely understood. To document the responses of cortical neurons in non-human primates to stimulation, we recorded single unit activity while delivering single-pulse stimulation via Utah arrays implanted in primary motor cortex of three macaque monkeys. Stimuli between 5-50 μA delivered to single channels reliably evoked spikes in neurons recorded throughout the array with delays of up to 12 milliseconds. ICMS pulses also induced a period of inhibition lasting up to 150 ms that typically followed the initial excitatory response. Higher current amplitudes led to a greater probability of evoking a spike and extended the duration of inhibition. The likelihood of evoking a spike in a neuron was dependent on the spontaneous firing rate as well as the delay between its most recent spike time and stimulus onset. Tonic repetitive stimulation between 2 and 20 Hz often modulated both the probability of evoking spikes and the duration of inhibition, although high frequency stimulation in particular was more likely to change both responses. On a trial-by-trial basis, whether a stimulus evoked a spike did not affect the subsequent inhibitory response; however, their changes over time were frequently positively or negatively correlated. Our results document the complex dynamics of cortical neural responses to electrical stimulation that need to be considered when utilizing ICMS for scientific and clinical applications.Significance statementIntracortical microstimulation (ICMS) is commonly used to probe the cortex, and previous studies have characterized the responses of single neurons to ICMS. However, these studies typically explored the averaged effects of ICMS throughout each experimental session, rather than by a trial-by-trial basis for each stimulation pulse. By shifting the approach, we explored the dependence of neural responses to ICMS on the spontaneous neural activity as well as the dynamics of responses over time due to repetitive stimulation. Our results highlight how the responses of neurons to ICMS are likely the result of interactions between local excitatory and inhibitory cortical circuits. These results will help inform the design of ICMS for both basic research and clinically relevant stimulation protocols.

Objective. Recent advances in neural decoding have accelerated the development of brain–computer interfaces aimed at assisting users with everyday tasks such as speaking, walking, and manipulating objects. However, current approaches for training neural decoders commonly require large quantities of labeled data, which can be laborious or infeasible to obtain in real-world settings. Alternatively, self-supervised models that share self-generated pseudo-labels between two data streams have shown exceptional performance on unlabeled audio and video data, but it remains unclear how well they extend to neural decoding. Approach. We learn neural decoders without labels by leveraging multiple simultaneously recorded data streams, including neural, kinematic, and physiological signals. Specifically, we apply cross-modal, self-supervised deep clustering to train decoders that can classify movements from brain recordings. After training, we then isolate the decoders for each input data stream and compare the accuracy of decoders trained using cross-modal deep clustering against supervised and unimodal, self-supervised models. Main results. We find that sharing pseudo-labels between two data streams during training substantially increases decoding performance compared to unimodal, self-supervised models, with accuracies approaching those of supervised decoders trained on labeled data. Next, we extend cross-modal decoder training to three or more modalities, achieving state-of-the-art neural decoding accuracy that matches or slightly exceeds the performance of supervised models. Significance. We demonstrate that cross-modal, self-supervised decoding can be applied to train neural decoders when few or no labels are available and extend the cross-modal framework to share information among three or more data streams, further improving self-supervised training.