A5031248654- Neural dynamics and brain function
- Congenital Diaphragmatic Hernia Studies
- Congenital heart defects research
- Functional Brain Connectivity Studies
- Craniofacial Disorders and Treatments
- Time Series Analysis and Forecasting
- Simulation Techniques and Applications
- Music and Audio Processing
- Neural Networks and Applications
- Scientific Computing and Data Management
- Gaussian Processes and Bayesian Inference
- Anomaly Detection Techniques and Applications
- Aerodynamics and Acoustics in Jet Flows
- Model Reduction and Neural Networks
- EEG and Brain-Computer Interfaces
- Speech and Audio Processing
- Speech Recognition and Synthesis
- Data Analysis with R
Tübingen AI Center
2023-2025
University of Tübingen
2023-2025
Bernstein Center for Computational Neuroscience Tübingen
2023-2025
Generative models are invaluable in many fields of science because their ability to capture high-dimensional and complicated distributions, such as photo-realistic images, protein structures, connectomes. How do we evaluate the samples these generate? This work aims provide an accessible entry point understanding popular notions statistical distances, requiring only foundational knowledge mathematics statistics. We focus on four commonly used distances representing different methodologies:...
Extracting the relationship between high-dimensional recordings of neural activity and complex behavior is a ubiquitous problem in systems neuroscience. Toward this goal, encoding decoding models attempt to infer conditional distribution given vice versa, while dimensionality reduction techniques aim extract interpretable low-dimensional representations. Variational autoencoders (VAEs) are flexible deep-learning commonly used embeddings or behavioral data. However, it challenging for VAEs...
Abstract In recent years, deep generative models have had a profound impact in engineering and sciences, revolutionizing domains such as image audio generation, well advancing our ability to model scientific data. particular, Denoising Diffusion Probabilistic Models (DDPMs) been shown accurately time series complex high-dimensional probability distributions. Experimental clinical neuroscience also stand benefit from this progress, since accurate modeling of neurophysiological series,...
Modern datasets in neuroscience enable unprecedented inquiries into the relationship between complex behaviors and activity of many simultaneously recorded neurons. While latent variable models can successfully extract low-dimensional embeddings from such recordings, using them to generate realistic spiking data, especially a behavior-dependent manner, still poses challenge. Here, we present Latent Diffusion for Neural Spiking data (LDNS), diffusion-based generative model with space: LDNS...
Scientists and engineers use simulators to model empirically observed phenomena. However, tuning the parameters of a simulator ensure its outputs match data presents significant challenge. Simulation-based inference (SBI) addresses this by enabling Bayesian for simulators, identifying that align with prior knowledge. Unlike traditional inference, SBI only needs access simulations from does not require evaluations likelihood-function. In addition, algorithms do gradients through simulator,...
Neonatal apneas and hypopneas present a serious risk for healthy infant development. Treating these adverse events requires frequent manual stimulation by skilled personnel, which can lead to alarm fatigue. This study aims develop validate an interpretable model that predict hypopneas. Automatically predicting before they occur would enable the use of methods automatic intervention. We propose neural additive individual occurrences neonatal apnea hypopnea apply it physiological dataset from...
Abstract Neonatal apneas and hypopneas present a serious risk for healthy infant development. Treating these adverse events requires frequent manual stimulation by skilled personnel, which can lead to alert fatigue. Automatically predicting before they occur would enable the use of methods automatic intervention. In this work, we propose neural additive model predict individual neonatal apnea hypopnea apply it physiological dataset from infants with Robin sequence at upper airway...
Scientific modeling applications often require estimating a distribution of parameters consistent with dataset observations - an inference task also known as source estimation. This problem can be ill-posed, however, since many different distributions might produce the same data-consistent simulations. To make principled choice among equally valid sources, we propose approach which targets maximum entropy distribution, i.e., prioritizes retaining much uncertainty possible. Our method is...