Embedded Files
For a complete list of publications please see here
Unsupervised Feature Selection Through Group Discovery
Unsupervised feature selection (FS) is essential for high-dimensional learning tasks where labels are not available. It helps reduce noise, improve generalization, and enhance interpretability. However, most existing unsupervised FS methods evaluate features in isolation, even though informative signals often emerge from groups of related features. For example, adjacent pixels, functionally connected brain regions, or correlated financial indicators tend to act together, making independent evaluation suboptimal. Although some methods attempt to capture group structure, they typically rely on predefined partitions or label supervision, limiting their applicability. We propose GroupFS, an end-to-end, fully differentiable framework that jointly discovers latent feature groups and selects the most informative groups among them, without relying on fixed a priori groups or label supervision. GroupFS enforces Laplacian smoothness on both feature and sample graphs and applies a group sparsity regularizer to learn a compact, structured representation. Across nine benchmarks spanning images, tabular data, and biological datasets, GroupFS consistently outperforms state-of-the-art unsupervised FS in clustering and selects groups of features that align with meaningful patterns.
Shira Lifshitz, Ofir Lindenbaum, Gal Mishne, Ron Meir and Hadas Benisty (Accepted to AAAI, preprint)
Unraveling Network Dynamics via RONI: Riemannian filtering Of Netowork Interactions
Functional connectivity (FC) is fundamentally non-stationary, undergoing continuous reconfigurations that track shifting behavioral and cognitive states. Despite the importance of these transitions, existing analytical frameworks struggle to reconcile the high-dimensional nature of these reconfigurations with the need for structured trajectories and mechanistic interpretability. Specifically, identifying the precise network components responsible for driving these dynamics remains a significant challenge. To bridge this gap, we introduce RONI (Riemannian filtering Of Network Interactions), a geometry-aware, unsupervised framework that treats time-varying FC as a signal evolving on the manifold of symmetric positive semidefinite (SPSD) matrices. By defining intrinsic filtering operators along the SPSD geodesics, RONI enables a principled multiresolution decomposition of FC dynamics directly on the manifold. This representation allows for the isolation of “dynamic drivers” - specific network elements that dominate coherent connectivity modes across distinct temporal scales. We demonstrate the versatility of RONI by applying it to a diverse array of large-scale neural recordings, including hippocampal electrophysiology data, cortex-wide and dendritic calcium imaging, and human EEG. Across these varied modalities and spatio-temporal scales, RONI identifies biologically meaningful sub-networks that shape the geometry of the connectivity trajectory, providing a unified, interpretable, and quantitative framework for studying the evolution of distributed neural interactions.
Yonatan Kleerekoper, Mohammad Kurtam, Yonatan Keselman, Shai Abramson, Dori Derdikman, Yitzhak Schiller, Simon Musall, and Hadas Benisty (Preprint)
VTA projections to M1 are essential for reorganization of layer 2-3 network dynamics underlying motor learning
The primary motor cortex (M1) is crucial for motor skill learning. Previous studies demonstrated that skill acquisition requires dopaminergic VTA (ventral-tegmental area) signaling in M1, however little is known regarding the effect of these inputs at the neuronal and network levels. We developed a novel geometric data analysis method for exploring the transformation of cellular dynamics induced by learning. Our findings demonstrate dopaminergic VTA-dependent formation of outcome signaling and new connectivity configuration of the layer 2-3 network, supporting reorganization of the M1 network for storing new motor skills.
Amir Ghanayim*, Hadas Benisty,*, Avigail Cohen-Rimon, Sivan Schwartz, Sally Dabdoob, Shira Lifshitz, Ronen Talmon and Jackie Schiller
Contextual Feature Selection with Conditional Stochastic Gates
We propose a novel architecture for learning the importance of each input variable for prediction of the target variable(s) as a function of a given context. Our extensive benchmark shows that c-STG improves feature selection, enhances prediction accuracy as well as model interpretability across multiple real-world domains.
Rapid fluctuations in functional connectivity of cortical networks encode spontaneous behavior
Experimental work across species has demonstrated that spontaneously generated behaviors are robustly coupled to variations in neural activity within the cerebral cortex. We used wide-field mesoscopic calcium imaging to monitor cortical dynamics in awake mice and developed an approach to quantify rapidly time-varying functional connectivity. Our approach generalizes the commonly used linear models linking behavior to neuronal activity: we formulate the modeling of behavior as a Taylor’s expansion where neuronal activity is the first order term and instantaneous correlations are the second order term. Using this model, we show that both terms are significant for modeling of spontaneous behaviors, meaning they are represented by fast changes in both the magnitude and correlational structure of cortical network activity. Moreover, the dynamic functional connectivity of mesoscale signals revealed subnetworks not predicted by traditional anatomical atlas-based parcellation of the cortex.
H. Benisty, A. Moberly, S. Lohani, D. Barson, R.R. Coifman, G. Mishne, J. Cardin, and M.J. Higley
Google Sites
Report abuse