Dmitri B. Chklovskii

This study evaluates the stability of connectome-constrained neural network models in identifying biologically plausible circuits, finding that task performance alone is insufficient for reliable mechanistic inference.

This paper develops a deep learning method to correct non-uniform milling artifacts in FIB-SEM images using unsupervised cross-plane image-to-image translation.

The paper presents the first 3D reconstruction of an insect compound eye using vEM, revealing cellular structure and specialized photoreceptors in the parasitoid wasp Megaphragma viggianii.

The paper develops an image-processing pipeline to restore VEM images with high fidelity while preserving relative contrast, tested on Microwasp head datasets.

This paper introduces a data-driven control framework to model neurons as feedback controllers, explaining various neurophysiological phenomena.

The paper proposes a mechanistic model of the Drosophila larval olfactory circuit that explains synaptic organization and function through Hebbian plasticity and activity correlations.

The paper proposes a linear discriminant analysis model of imbalanced associative learning in the mushroom body compartment, explaining how dopamine neurons modulate synaptic plasticity.

The paper introduces two biologically plausible algorithms for segmenting time series based on underlying dynamics, showing good performance on autoregressive models and voice recordings.

The paper investigates representational drift in neural networks, showing that noisy synaptic updates cause coordinated random walks in receptive fields while maintaining population code stability.

CaImAn, an open-source library, automates motion correction, source extraction, deconvolution and cross-session registration for calcium imaging, detecting neurons near human level (F1 0.75 batch, 0.76 online).

The paper shows that similarity-preserving neural networks develop localized receptive fields that tile manifolds, mirroring brain representations like place cells.

OnACID processes streaming calcium imaging data in real time by combining online dictionary learning with motion correction, source extraction, and activity deconvolution, tracking hundreds of neurons for closed-loop experiments.

The paper proposes that the PN-KC network in insect olfaction implements an online clustering algorithm derived from the k-means cost function, explaining key features of the circuit.