A5023054628- Bayesian Modeling and Causal Inference
- Machine Learning and Algorithms
- Domain Adaptation and Few-Shot Learning
- Machine Learning and Data Classification
- Medical Imaging and Analysis
- Acute Ischemic Stroke Management
- Explainable Artificial Intelligence (XAI)
- Reinforcement Learning in Robotics
- Complex Systems and Decision Making
- Advanced Graph Neural Networks
- Natural Language Processing Techniques
- Cerebrovascular and Carotid Artery Diseases
- Statistical Methods and Inference
- Machine Learning in Healthcare
- Cellular Mechanics and Interactions
- Game Theory and Applications
- Topic Modeling
- 3D Printing in Biomedical Research
- Wound Healing and Treatments
- Ethics in Clinical Research
ETH Zurich
2020-2023
Scherrer (Switzerland)
2020
Institute for Biomedical Engineering
2020
University of Zurich
2020
While Large Language Models require more and data to train scale, rather than looking for any acquire, we should consider what types of tasks are likely benefit from scaling. We be intentional in our acquisition. argue that the topology itself informs which prioritize scaling, shapes development next generation compute paradigms where scaling is inefficient, or even insufficient.
Discovering causal structures from data is a challenging inference problem of fundamental importance in all areas science. The appealing properties neural networks have recently led to surge interest differentiable network-based methods for learning data. So far, discovery has focused on static datasets observational or fixed interventional origin. In this work, we introduce an active intervention targeting (AIT) method which enables quick identification the underlying structure...
To compare the segmentation and detection performance of a deep learning model trained on database human-labeled clinical stroke lesions diffusion-weighted (DW) images to same enhanced with synthetic lesions.In this institutional review board-approved study, 962 cases (mean patient age ± standard deviation, 65 years 17; 255 male patients; 449 scans DW positive lesions) normal 2027 patients age, 38 24; 1088 female patients) were used. Brain volumes produced by warping relative signal increase...
Transformers have become the dominant model in deep learning, but reason for their superior performance is poorly understood. Here, we hypothesize that strong of stems from an architectural bias towards mesa-optimization, a learned process running within forward pass consisting following two steps: (i) construction internal learning objective, and (ii) its corresponding solution found through optimization. To test this hypothesis, reverse-engineer series autoregressive trained on simple...
Learning the causal structure that underlies data is a crucial step towards robust real-world decision making. The majority of existing work in inference focuses on determining single directed acyclic graph (DAG) or Markov equivalence class thereof. However, aspect to acting intelligently upon knowledge about which has been inferred from finite demands reasoning its uncertainty. For instance, planning interventions find out more mechanisms govern our requires quantifying epistemic...
Learning models that offer robust out-of-distribution generalization and fast adaptation is a key challenge in modern machine learning. Modelling causal structure into neural networks holds the promise to accomplish zero few-shot adaptation. Recent advances differentiable discovery have proposed factorize data generating process set of modules, i.e. one module for conditional distribution every variable where only parents are used as predictors. Such modular decomposition knowledge enables...
Self-interested individuals often fail to cooperate, posing a fundamental challenge for multi-agent learning. How can we achieve cooperation among self-interested, independent learning agents? Promising recent work has shown that in certain tasks be established between learning-aware agents who model the dynamics of each other. Here, present first unbiased, higher-derivative-free policy gradient algorithm reinforcement learning, which takes into account other are themselves through trial and...
Causal discovery serves a pivotal role in mitigating model uncertainty through recovering the underlying causal mechanisms among variables. In many practical domains, such as healthcare, access to data gathered by individual entities is limited, primarily for privacy and regulatory constraints. However, majority of existing methods require be available centralized location. response, researchers have introduced federated discovery. While previous consider distributed observational data,...
Human skin equivalents (HSEs) serve as important tools for mechanistic studies with human cells, drug discovery, pre-clinical applications in the field of tissue engineering and transplantation on defects. Besides cellular extracellular matrix (ECM) components used HSEs, physical constraints applied scaffold during HSEs maturation influence organization, functionality, homogeneity. In this study, we introduce a 3D-printed culture insert that exposes bi-layered to static radial constraint...
Introduction: The application of deep learning to stroke image analysis (and medical images in general) faces two major challenges: first, it requires a large number train, which is difficult obtain. Second, the accurate outlining infarcts tedious, high level expertise, subjective and error prone. purpose this work was produce set diffusion-weighted (DWI) with perfectly defined realistic-appearing synthetic acute lesions compare segmentation performance neural network trained on these DWI...
Inferring causal structure from data is a challenging task of fundamental importance in science. Observational are often insufficient to identify system's uniquely. While conducting interventions (i.e., experiments) can improve the identifiability, such samples usually and expensive obtain. Hence, experimental design approaches for discovery aim minimize number by estimating most informative intervention target. In this work, we propose novel Gradient-based Intervention Targeting method,...