Do your eyes glaze over when looking at a long list of annual health insurance enrollment options – or maybe while you’re trying to calculate how much you owe the IRS? You might be wondering the same thing we are: Where’s the guidebook for all of this grown-up stuff? Whether opening a bank account, refinancing student loans, or purchasing car insurance (...um, can we just roll the dice without it?), we’re just as confused as you are. Enter: “Grown-Up Stuff: How to Adult” a podcast dedicated ...
…
continue reading
Sisällön tarjoaa Oxford University. Oxford University tai sen podcast-alustan kumppani lataa ja toimittaa kaiken podcast-sisällön, mukaan lukien jaksot, grafiikat ja podcast-kuvaukset. Jos uskot jonkun käyttävän tekijänoikeudella suojattua teostasi ilman lupaasi, voit seurata tässä https://fi.player.fm/legal kuvattua prosessia.
Player FM - Podcast-sovellus
Siirry offline-tilaan Player FM avulla!
Siirry offline-tilaan Player FM avulla!
Neural Networks and Deep Kernel Shaping
Manage episode 324917919 series 1610930
Sisällön tarjoaa Oxford University. Oxford University tai sen podcast-alustan kumppani lataa ja toimittaa kaiken podcast-sisällön, mukaan lukien jaksot, grafiikat ja podcast-kuvaukset. Jos uskot jonkun käyttävän tekijänoikeudella suojattua teostasi ilman lupaasi, voit seurata tässä https://fi.player.fm/legal kuvattua prosessia.
Rapid training of deep neural networks without skip connections or normalization layers using Deep Kernel Shaping. Using an extended and formalized version of the Q/C map analysis of Pool et al. (2016), along with Neural Tangent Kernel theory, we identify the main pathologies present in deep networks that prevent them from training fast and generalizing to unseen data, and show how these can be avoided by carefully controlling the "shape" of the network's initialization-time kernel function. We then develop a method called Deep Kernel Shaping (DKS), which accomplishes this using a combination of precise parameter initialization, activation function transformations, and small architectural tweaks, all of which preserve the model class. In our experiments we show that DKS enables SGD training of residual networks without normalization layers on Imagenet and CIFAR-10 classification tasks at speeds comparable to standard ResNetV2 and Wide-ResNet models, with only a small decrease in generalization performance. And when using K-FAC as the optimizer, we achieve similar results for networks without skip connections. Our results apply for a large variety of activation functions, including those which traditionally perform very badly, such as the logistic sigmoid. In addition to DKS, we contribute a detailed analysis of skip connections, normalization layers, special activation functions like RELU and SELU, and various initialization schemes, explaining their effectiveness as alternative (and ultimately incomplete) ways of "shaping" the network's initialization-time kernel.
…
continue reading
51 jaksoa
Manage episode 324917919 series 1610930
Sisällön tarjoaa Oxford University. Oxford University tai sen podcast-alustan kumppani lataa ja toimittaa kaiken podcast-sisällön, mukaan lukien jaksot, grafiikat ja podcast-kuvaukset. Jos uskot jonkun käyttävän tekijänoikeudella suojattua teostasi ilman lupaasi, voit seurata tässä https://fi.player.fm/legal kuvattua prosessia.
Rapid training of deep neural networks without skip connections or normalization layers using Deep Kernel Shaping. Using an extended and formalized version of the Q/C map analysis of Pool et al. (2016), along with Neural Tangent Kernel theory, we identify the main pathologies present in deep networks that prevent them from training fast and generalizing to unseen data, and show how these can be avoided by carefully controlling the "shape" of the network's initialization-time kernel function. We then develop a method called Deep Kernel Shaping (DKS), which accomplishes this using a combination of precise parameter initialization, activation function transformations, and small architectural tweaks, all of which preserve the model class. In our experiments we show that DKS enables SGD training of residual networks without normalization layers on Imagenet and CIFAR-10 classification tasks at speeds comparable to standard ResNetV2 and Wide-ResNet models, with only a small decrease in generalization performance. And when using K-FAC as the optimizer, we achieve similar results for networks without skip connections. Our results apply for a large variety of activation functions, including those which traditionally perform very badly, such as the logistic sigmoid. In addition to DKS, we contribute a detailed analysis of skip connections, normalization layers, special activation functions like RELU and SELU, and various initialization schemes, explaining their effectiveness as alternative (and ultimately incomplete) ways of "shaping" the network's initialization-time kernel.
…
continue reading
51 jaksoa
Kaikki jaksot
×Tervetuloa Player FM:n!
Player FM skannaa verkkoa löytääkseen korkealaatuisia podcasteja, joista voit nauttia juuri nyt. Se on paras podcast-sovellus ja toimii Androidilla, iPhonela, ja verkossa. Rekisteröidy sykronoidaksesi tilaukset laitteiden välillä.