PinnedPublished inTDS ArchiveTrain ImageNet without Hyperparameters with Automatic Gradient DescentTowards architecture-aware optimisationApr 19, 2023Apr 19, 2023
Published inTDS ArchiveWhat can flatness teach us: understanding generalisation in Deep Neural NetworksThis is the third post in a series summarising work that seeks to provide a theory of generalisation in Deep Neural Networks (DNNs)…Mar 29, 2021Mar 29, 2021
Published inTDS ArchiveDeep Neural Networks are biased, at initialisation, towards simple functionsAnd why is this a very important step in understanding why they work?Jan 1, 2021Jan 1, 2021
Published inTDS ArchiveNeural networks are fundamentally BayesianStochastic Gradient Descent approximates Bayesian samplingDec 19, 20202Dec 19, 20202