J. Cunningham
5 Papers
2 Citations
J. Cunningham is an academic researcher. The author has contributed to research in topics: Computer science & Biology. The author has an hindex of 2, co-authored 5 publications.
Chat about Author
Papers
Practical and Asymptotically Exact Conditional Sampling in Diffusion Models
TL;DR: The Twisted Diffusion Sampler (TDS) as mentioned in this paper is a sequential Monte Carlo (SMC) algorithm that targets the conditional distributions of diffusion models, which can provide exact samples for a broad range of conditional distributions without requiring task-specific training.
34
Pathologies of Predictive Diversity in Deep Ensembles
TL;DR: The authors showed that diversity-encouraging regularizers hurt the performance of high-capacity deep ensembles used for classification, while discouraging predictive diversity can be beneficial, which suggests that the best strategy for deep ensemble is utilizing more accurate, but likely less diverse, component models.
4
The Best Deep Ensembles Sacrifice Predictive Diversity
TL;DR: The authors investigated the tradeoff between diversity and individual model performance and found that encouraging diversity during training almost always yields worse ensembles, and that the Jensen gap is a natural measure of diversity for both the mean squared error and cross entropy loss functions.
3
Lightning Pose: improved animal pose estimation via semi-supervised learning, Bayesian ensembling, and cloud-native open-source tools
Dan Biderman,Matthew R Whiteway,Cole Lincoln Hurwitz,Ankit Vishnubhotla,Federico Pedraja,Michael M Schartner,Julia M. Huntenburg,Anup Khanal,Guido T. Meijer,Jean-Paul Noel,Alejandro Pan-Vazquez,Karolina Socha,Anne E Urai,J. Cunningham,Nathaniel B. Sawtell,Liam Paninski +15 more
TL;DR: In this article , a semi-supervised approach is proposed that leverages the spatio-temporal statistics of unlabeled videos in two different ways: first, they introduce unsupervised training objectives that penalize the network whenever its predictions violate smoothness of physical motion, multiple-view geometry, or depart from a low-dimensional subspace of plausible body configurations.
Simple decoding of behavior from a complicated neural manifold
TL;DR: In this paper , a trajectory-centric approach is proposed for decoding motor-cortical neural trajectories, where each neural trajectory has a corresponding behavioral trajectory, allowing straightforward but highly nonlinear decoding.