Epistemic Autonomy: Self-supervised Learning in the Mammalian Hippocampus.
Trends in cognitive sciences
Biological cognition is based on the ability to autonomously acquire knowledge, or epistemic autonomy. Such self-supervision is largely absent in artificial neural networks (ANN) because they depend on externally set learning criteria. Yet training ANN using error backpropagation has created the current revolution in artificial intelligence, raising the question of whether the epistemic autonomy displayed in biological cognition can be achieved with error backpropagation-based learning. We present evidence suggesting that the entorhinal-hippocampal complex combines epistemic autonomy with error backpropagation. Specifically, we propose that the hippocampus minimizes the error between its input and output signals through a modulatory counter-current inhibitory network. We further discuss the computational emulation of this principle and analyze it in the context of autonomous cognitive systems.
View details for DOI 10.1016/j.tics.2021.03.016
View details for PubMedID 33906817
Entorhinal mismatch: A model of self-supervised learning in the hippocampus.
2021; 24 (4): 102364
The hippocampal formation displays a wide range of physiological responses to different spatial manipulations of the environment. However, very few attempts have been made to identify core computational principles underlying those hippocampal responses. Here, we capitalize on the observation that the entorhinal-hippocampal complex (EHC) forms a closed loop and projects inhibitory signals "countercurrent" to the trisynaptic pathway to build a self-supervised model that learns to reconstruct its own inputs by error backpropagation. The EHC is then abstracted as an autoencoder, with the hidden layers acting as an information bottleneck. With the inputs mimicking the firing activity of lateral and medial entorhinal cells, our model is shown to generate place cells and to respond to environmental manipulations as observed in rodent experiments. Altogether, we propose that the hippocampus builds conjunctive compressed representations of the environment by learning to reconstruct its own entorhinal inputs via gradient descent.
View details for DOI 10.1016/j.isci.2021.102364
View details for PubMedID 33997671
Maximally selective single-cell target for circuit control in epilepsy models.
Neurological and psychiatric disorders are associated with pathological neural dynamics. The fundamental connectivity patterns of cell-cell communication networks that enable pathological dynamics to emerge remain unknown. Here, we studied epileptic circuits using a newly developed computational pipeline that leveraged single-cell calcium imaging of larval zebrafish and chronically epileptic mice, biologically constrained effective connectivity modeling, and higher-order motif-focused network analysis. We uncovered a novel functional cell type that preferentially emerged in the preseizure state, the superhub, that was unusually richly connected to the rest of the network through feedforward motifs, critically enhancing downstream excitation. Perturbation simulations indicated that disconnecting superhubs was significantly more effective in stabilizing epileptic circuits than disconnecting hub cells that were defined traditionally by connection count. In the dentate gyrus of chronically epileptic mice, superhubs were predominately modeled adult-born granule cells. Collectively, these results predict a new maximally selective and minimally invasive cellular target for seizure control.
View details for DOI 10.1016/j.neuron.2021.06.007
View details for PubMedID 34197732
- Data-Driven Modeling of Normal and Pathological Oscillations in the Hippocampus MULTISCALE MODELS OF BRAIN DISORDERS 2019; 13: 185–92
- Network Models of Epilepsy-Related Pathological Structural and Functional Alterations in the Dentate Gyrus REWIRING BRAIN: A COMPUTATIONAL APPROACH TO STRUCTURAL PLASTICITY IN THE ADULT BRAIN 2017: 485–503
Interneuronal mechanisms of hippocampal theta oscillations in a full-scale model of the rodent CA1 circuit.
The hippocampal theta rhythm plays important roles in information processing; however, the mechanisms of its generation are not well understood. We developed a data-driven, supercomputer-based, full-scale (1:1) model of the rodent CA1 area and studied its interneurons during theta oscillations. Theta rhythm with phase-locked gamma oscillations and phase-preferential discharges of distinct interneuronal types spontaneously emerged from the isolated CA1 circuit without rhythmic inputs. Perturbation experiments identified parvalbumin-expressing interneurons and neurogliaform cells, as well as interneuronal diversity itself, as important factors in theta generation. These simulations reveal new insights into the spatiotemporal organization of the CA1 circuit during theta oscillations.
View details for DOI 10.7554/eLife.18566
View details for PubMedID 28009257
View details for PubMedCentralID PMC5313080