Published inTDS ArchiveA Search for Efficient Meta-Learning: MAMLs, Reptiles, and Related SpeciesIt’s an oft-lamented fact that the capabilities of modern machine learning tend to be narrow and brittle: while a given technique can be…Sep 27, 2020A response icon1Sep 27, 2020A response icon1
Published inTDS ArchiveConditional Love: The Rise of Renormalization Techniques for Neural Network ConditioningConditional renormalization is an oft-unsung technique powering many recent ML successes; how does it work and where did the idea come…Sep 8, 2019A response icon2Sep 8, 2019A response icon2
Published inTDS ArchiveIt’s Only Natural: An Excessively Deep Dive Into Natural Gradient OptimizationThis post gives an intuitive explanation of an approach called Natural Gradient, an elegant way to dynamically adjust gradient step size.Mar 9, 2019A response icon9Mar 9, 2019A response icon9
Published inTDS ArchiveGenerating, With Style: The Mechanics Behind NVIDIA’s Highly Realistic GAN ImagesA new architecture, designed to leverage noise and allow for precise configuration of global features, is generating impressive resultsJan 21, 2019A response icon1Jan 21, 2019A response icon1
Published inTDS ArchiveA Tale of Two Convolutions: Differing Design Paradigms for Graph Neural Networks“Graph” is a one of those terms that’s fallen prey to natural language’s tendency to be less precise than its mathematical counterpart.Dec 27, 2018A response icon4Dec 27, 2018A response icon4
Published inTDS ArchiveConvolution: An Exploration of a Familiar Operator’s Deeper RootsIn the world of modern machine learning, the convolution operator occupies the strange position: it’s both trivially familiar to anyone…Oct 1, 2018A response icon5Oct 1, 2018A response icon5
Published inTDS ArchiveThe Pursuit of (Robotic) Happiness: How TRPO and PPO Stabilize Policy Gradient MethodsReinforcement Learning strikes me as the wild west of machine learning right now: a place full of drama and progress, with dreams of grand…Jul 9, 2018A response icon5Jul 9, 2018A response icon5
Published inTDS ArchiveWith Great Power Comes Poor Latent Codes: Representation Learning in VAEs (Pt. 2)(If you haven’t done so yet, I recommend going back and reading Part 1 of this series on VAE failure modes; I spent more time there…May 7, 2018A response icon7May 7, 2018A response icon7
Published inTDS ArchiveWhat a Disentangled Net We Weave: Representation Learning in VAEs (Pt. 1)It’s a truth universally acknowledged: that data not in possession of labels must be in want of unsupervised learning.Apr 15, 2018A response icon7Apr 15, 2018A response icon7
Published inTDS ArchiveLearning About Algorithms That Learn to LearnThe premise of meta learning was an intoxicating one to me, when I first of heard it: the project of building machines that are not only…Mar 24, 2018A response icon2Mar 24, 2018A response icon2