INNF+ 2021 ICML Workshop on Invertible Neural Networks, Normalizing Flows, and Explicit Likelihood Models --> --> December 2, 2018 Le 1000 Conference Center 1000 Rue de la Gauchetière Ouest Montréal, QC H3B 0A2, Canada --> --> --> Accepted Papers --> .paper {margin-bottom:10pt; border-bottom: solid silver 1px;} --> Block Neural Autoregressive Flow Nicola De Cao , Wilker Aziz and Ivan Titov . --> Understanding Event-Generation Networks via Uncertainties Abstract Generative models and normalizing flow based

Veronika Koren, Ph.D. Computational Neuroscientist Home Contact me Firing rates and representational error in efficient spiking networks are bounded by design Authors: Matin Urdu, Gabriel Matías Lorenz, Ching-Peng Huang, Stefano Panzeri, Veronika Koren Accepted at the International Conference on Artificial Neural Networks 2025 (ICANN 2025) Disclaimer: This version of the contribution has been accepted for publication, after peer review but is not the Version of Record and does not reflect post-acceptance

I’ve trained this open-source neural network framework on a variety of datasets, including recipes, Pokemon, knock-knock jokes, pick up lines, and D&D spells. This time, the dataset was a list of almost 2000 car models, helpfully provided by readers Anya and Dan. Can the neural network learn to suggest plausible car models? (Spoiler: Well, kinda

A recap of NEXT24 through the lens of networks and their rise, from computers to the brain, from democracy to energy

We study the high-dimensional training dynamics of a shallow neural network with quadratic activation in a teacher-student setup. We focus on the extensive-width regime, where the teacher and

Wherein the Practice of Ensembling Neural Networks Is Recounted, With Dropout Presented as an Implicit Ensemble Approximating a Deep Gaussian Process, and BatchEnsemble Tricks Are Adopted Due to GPU Constraints

Deep neural network models of sensory systems are often proposed to learn representational transformations with invariances like those in the brain. To reveal these invariances, we generated ‘model metamers’, stimuli whose activations within a model stage are matched to those of a natural stimulus. Metamers for state-of-the-art supervised and unsupervised neural network models of vision and audition were often completely unrecognizable to humans when generated from late model stages, suggesting

CNNs for deep learning

This tutorial shows you how to use multiple GPUs to train your TensorFlow neural networks. You'll learn how to use data parallelism to train your models on

A memory framework combining short-term and long-term memory in neural networks improves long-sequence modeling efficiency and performance