Yizi Zhang

All Publications

• Exploiting correlations across trials and behavioral sessions to improve neural decoding. Neuron Zhang, Y., Lyu, H., Hurwitz, C., Wang, S., Findling, C., Wang, Y., Hubert, F., Pouget, A., Varol, E., Paninski, L. 2026; 114 (3): 536-551.e11

  Abstract

  Traditional neural decoders link neural activity to behavior within single trials of a session, overlooking correlations across trials and sessions. However, animals show similar neural patterns when performing the same task, and their behaviors are influenced by prior experiences. To capture these dependencies, we introduce two complementary models: a multi-session reduced-rank regression model that shares behaviorally relevant neural structure across sessions and a multi-session state-space model that captures behavioral structure across trials and sessions. On 433 sessions spanning 270 brain regions in the International Brain Laboratory (IBL) mouse Neuropixels dataset, our decoders outperform traditional approaches on four behaviors, with results generalizing across datasets, species, and tasks. Unlike deep learning methods, our models are efficient and interpretable, providing low-dimensional neural representations, task-related single-neuron contributions, and brain-wide timescales of neural activation.

  View details for DOI 10.1016/j.neuron.2025.10.026

  View details for PubMedID 41308644

  View details for PubMedCentralID PMC12695064

• Brain-wide representations of prior information in mouse decision-making. Nature Findling, C., Hubert, F., Acerbi, L., Benson, B., Benson, J., Birman, D., Bonacchi, N., Buchanan, E. K., Bruijns, S., Carandini, M., Catarino, J. A., Chapuis, G. A., Churchland, A. K., Dan, Y., Davatolhagh, F., DeWitt, E. E., Engel, T. A., Fabbri, M., Faulkner, M. A., Fiete, I. R., Freitas-Silva, L., Gerçek, B., Harris, K. D., Häusser, M., Hofer, S. B., Hu, F., Huntenburg, J. M., Khanal, A., Krasniak, C., Langdon, C., Langfield, C. A., Latham, P. E., Lau, P. Y., Mainen, Z., Meijer, G. T., Miska, N. J., Mrsic-Flogel, T. D., Noel, J. P., Nylund, K., Pan-Vazquez, A., Paninski, L., Pillow, J., Rossant, C., Roth, N., Schaeffer, R., Schartner, M., Shi, Y., Socha, K. Z., Steinmetz, N. A., Svoboda, K., Tessereau, C., Urai, A. E., Wells, M. J., West, S. J., Whiteway, M. R., Winter, O., Witten, I. B., Zador, A., Zhang, Y., Dayan, P., Pouget, A. 2025; 645 (8079): 192-200

  Abstract

  The neural representations of prior information about the state of the world are poorly understood1. Here, to investigate them, we examined brain-wide Neuropixels recordings and widefield calcium imaging collected by the International Brain Laboratory. Mice were trained to indicate the location of a visual grating stimulus, which appeared on the left or right with a prior probability alternating between 0.2 and 0.8 in blocks of variable length. We found that mice estimate this prior probability and thereby improve their decision accuracy. Furthermore, we report that this subjective prior is encoded in at least 20% to 30% of brain regions that, notably, span all levels of processing, from early sensory areas (the lateral geniculate nucleus and primary visual cortex) to motor regions (secondary and primary motor cortex and gigantocellular reticular nucleus) and high-level cortical regions (the dorsal anterior cingulate area and ventrolateral orbitofrontal cortex). This widespread representation of the prior is consistent with a neural model of Bayesian inference involving loops between areas, as opposed to a model in which the prior is incorporated only in decision-making areas. This study offers a brain-wide perspective on prior encoding at cellular resolution, underscoring the importance of using large-scale recordings on a single standardized task.

  View details for DOI 10.1038/s41586-025-09226-1

  View details for PubMedID 40903597

  View details for PubMedCentralID PMC12408363

• Reproducibility of in vivo electrophysiological measurements in mice. eLife Banga, K., Benson, J., Bhagat, J., Biderman, D., Birman, D., Bonacchi, N., Bruijns, S. A., Buchanan, K., Campbell, R. A., Carandini, M., Chapuis, G. A., Churchland, A. K., Davatolhagh, M. F., Lee, H. D., Faulkner, M., Gerçek, B., Hu, F., Huntenburg, J., Hurwitz, C. L., Khanal, A., Krasniak, C., Lau, P., Langfield, C., Mackenzie, N., Meijer, G. T., Miska, N. J., Mohammadi, Z., Noel, J. P., Paninski, L., Pan-Vazquez, A., Rossant, C., Roth, N., Schartner, M., Socha, K. Z., Steinmetz, N. A., Svoboda, K., Taheri, M., Urai, A. E., Wang, S., Wells, M., West, S. J., Whiteway, M. R., Winter, O., Witten, I. B., Zhang, Y. 2025; 13

  Abstract

  Understanding brain function relies on the collective work of many labs generating reproducible results. However, reproducibility has not been systematically assessed within the context of electrophysiological recordings during cognitive behaviors. To address this, we formed a multi-lab collaboration using a shared, open-source behavioral task and experimental apparatus. Experimenters in 10 laboratories repeatedly targeted Neuropixels probes to the same location (spanning secondary visual areas, hippocampus, and thalamus) in mice making decisions; this generated a total of 121 experimental replicates, a unique dataset for evaluating reproducibility of electrophysiology experiments. Despite standardizing both behavioral and electrophysiological procedures, some experimental outcomes were highly variable. A closer analysis uncovered that variability in electrode targeting hindered reproducibility, as did the limited statistical power of some routinely used electrophysiological analyses, such as single-neuron tests of modulation by individual task parameters. Reproducibility was enhanced by histological and electrophysiological quality-control criteria. Our observations suggest that data from systems neuroscience is vulnerable to a lack of reproducibility, but that across-lab standardization, including metrics we propose, can serve to mitigate this.

  View details for DOI 10.7554/eLife.100840

  View details for PubMedID 40354112

• Neural Encoding and Decoding at Scale Zhang, Y., Wang, Y., Azabou, M., Andre, A., Wang, Z., Lyu, H., Dyer, E., Paninski, L., Hurwitz, C., Int Brain Lab edited by Singh, A., Fazel, M., Hsu, D., Lacoste-Julien, S., Berkenkamp, F., Maharaj, T., Wagstaff, K., Zhu, J. JMLR-JOURNAL MACHINE LEARNING RESEARCH. 2025: 76175-76192

  View details for Web of Science ID 001693172600041

• Towards a "universal translator" for neural dynamics at single-cell, single-spike resolution. ArXiv Zhang, Y., Wang, Y., Beneto, D. J., Wang, Z., Azabou, M., Richards, B., Winter, O., International Brain Laboratory, Dyer, E., Paninski, L., Hurwitz, C. 2024

  Abstract

  Neuroscience research has made immense progress over the last decade, but our understanding of the brain remains fragmented and piecemeal: the dream of probing an arbitrary brain region and automatically reading out the information encoded in its neural activity remains out of reach. In this work, we build towards a first foundation model for neural spiking data that can solve a diverse set of tasks across multiple brain areas. We introduce a novel self-supervised modeling approach for population activity in which the model alternates between masking out and reconstructing neural activity across different time steps, neurons, and brain regions. To evaluate our approach, we design unsupervised and supervised prediction tasks using the International Brain Laboratory repeated site dataset, which is comprised of Neuropixels recordings targeting the same brain locations across 48 animals and experimental sessions. The prediction tasks include single-neuron and region-level activity prediction, forward prediction, and behavior decoding. We demonstrate that our multi-task-masking (MtM) approach significantly improves the performance of current state-of-the-art population models and enables multi-task learning. We also show that by training on multiple animals, we can improve the generalization ability of the model to unseen animals, paving the way for a foundation model of the brain at single-cell, single-spike resolution. Project page and code: https://ibl-mtm.github.io/.

  View details for PubMedID 39108295

• Rhesus infant nervous temperament predicts peri-adolescent central amygdala metabolism & behavioral inhibition measured by a machine-learning approach. Translational psychiatry Holley, D., Campos, L. J., Drzewiecki, C. M., Zhang, Y., Capitanio, J. P., Fox, A. S. 2024; 14 (1): 148

  Abstract

  Anxiety disorders affect millions of people worldwide and impair health, happiness, and productivity on a massive scale. Developmental research points to a connection between early-life behavioral inhibition and the eventual development of these disorders. Our group has previously shown that measures of behavioral inhibition in young rhesus monkeys (Macaca mulatta) predict anxiety-like behavior later in life. In recent years, clinical and basic researchers have implicated the central extended amygdala (EAc)-a neuroanatomical concept that includes the central nucleus of the amygdala (Ce) and the bed nucleus of the stria terminalis (BST)-as a key neural substrate for the expression of anxious and inhibited behavior. An improved understanding of how early-life behavioral inhibition relates to an increased lifetime risk of anxiety disorders-and how this relationship is mediated by alterations in the EAc-could lead to improved treatments and preventive strategies. In this study, we explored the relationships between infant behavioral inhibition and peri-adolescent defensive behavior and brain metabolism in 18 female rhesus monkeys. We coupled a mildly threatening behavioral assay with concurrent multimodal neuroimaging, and related those findings to various measures of infant temperament. To score the behavioral assay, we developed and validated UC-Freeze, a semi-automated machine-learning (ML) tool that uses unsupervised clustering to quantify freezing. Consistent with previous work, we found that heightened Ce metabolism predicted elevated defensive behavior (i.e., more freezing) in the presence of an unfamiliar human intruder. Although we found no link between infant-inhibited temperament and peri-adolescent EAc metabolism or defensive behavior, we did identify infant nervous temperament as a significant predictor of peri-adolescent defensive behavior. Our findings suggest a connection between infant nervous temperament and the eventual development of anxiety and depressive disorders. Moreover, our approach highlights the potential for ML tools to augment existing behavioral neuroscience methods.

  View details for DOI 10.1038/s41398-024-02858-3

  View details for PubMedID 38490997

  View details for PubMedCentralID PMC10943234

• Predicting rare outcomes in abdominal wall reconstruction using image-based deep learning models. Surgery Ayuso, S. A., Elhage, S. A., Zhang, Y., Aladegbami, B. G., Gersin, K. S., Fischer, J. P., Augenstein, V. A., Colavita, P. D., Heniford, B. T. 2023; 173 (3): 748-755

  Abstract

  Deep learning models with imbalanced data sets are a challenge in the fields of artificial intelligence and surgery. The aim of this study was to develop and compare deep learning models that predict rare but devastating postoperative complications after abdominal wall reconstruction.A prospectively maintained institutional database was used to identify abdominal wall reconstruction patients with preoperative computed tomography scans. Conventional deep learning models were developed using an 8-layer convolutional neural network and a 2-class training system (ie, learns negative and positive outcomes). Conventional deep learning models were compared to deep learning models that were developed using a generative adversarial network anomaly framework, which uses image augmentation and anomaly detection. The primary outcomes were receiver operating characteristic values for predicting mesh infection and pulmonary failure.Computed tomography scans from 510 patients were used with a total of 10,004 images. Mesh infection and pulmonary failure occurred in 3.7% and 5.6% of patients, respectively. The conventional deep learning models were less effective than generative adversarial network anomaly for predicting mesh infection (receiver operating characteristic 0.61 vs 0.73, P < .01) and pulmonary failure (receiver operating characteristic 0.59 vs 0.70, P < .01). Although the conventional deep learning models had higher accuracies/specificities for predicting mesh infection (0.93 vs 0.78, P < .01/.96 vs .78, P < .01) and pulmonary failure (0.88 vs 0.68, P < .01/.92 vs .67, P < .01), they were substantially compromised by decreased model sensitivity (0.25 vs 0.68, P < .01/.27 vs .73, P < .01).Compared to conventional deep learning models, generative adversarial network anomaly deep learning models showed improved performance on imbalanced data sets, predominantly by increasing model sensitivity. Understanding patients who are at risk for rare but devastating postoperative complications can improve risk stratification, resource utilization, and the consent process.

  View details for DOI 10.1016/j.surg.2022.06.048

  View details for PubMedID 36229252

• Bypassing spike sorting: Density-based decoding using spike localization from dense multielectrode probes Zhang, Y., He, T., Boussard, J., Windolf, C., Winter, O., Trautmann, E., Roth, N., Barrell, H., Churchland, M., Steinmetz, N. A., Varol, E., Hurwitz, C., Paninski, L., Int Brain Lab edited by Oh, A., Neumann, T., Globerson, A., Saenko, K., Hardt, M., Levine, S. NEURAL INFORMATION PROCESSING SYSTEMS (NIPS). 2023

  View details for Web of Science ID 001220600001005