Timezone: »

Learned Optimizers that Scale and Generalize
Olga Wichrowska · Niru Maheswaranathan · Matthew Hoffman · Sergio Gómez Colmenarejo · Misha Denil · Nando de Freitas · Jascha Sohl-Dickstein

Mon Aug 07 12:33 AM -- 12:51 AM (PDT) @ Darling Harbour Theatre

Learning to learn has emerged as an important direction for achieving artificial intelligence. Two of the primary barriers to its adoption are an inability to scale to larger problems and a limited ability to generalize to new tasks. We introduce a learned gradient descent optimizer that generalizes well to new tasks, and which has significantly reduced memory and computation overhead. We achieve this by introducing a novel hierarchical RNN architecture, with minimal per-parameter overhead, augmented with additional architectural features that mirror the known structure of optimization tasks. We also develop a meta-training ensemble of small, diverse, optimization tasks capturing common properties of loss landscapes. The optimizer learns to outperform RMSProp/ADAM on problems in this corpus. More importantly, it performs comparably or better when applied to small convolutional neural networks, despite seeing no neural networks in its meta-training set. Finally, it generalizes to train Inception V3 and ResNet V2 architectures on the ImageNet dataset for thousands of steps, optimization problems that are of a vastly different scale than those it was trained on.

Author Information

Olga Wichrowska (Google Brain)
Niru Maheswaranathan (Stanford University)
Matthew Hoffman (DeepMind)
Sergio Gómez Colmenarejo (Google DeepMind)
Misha Denil (University of Oxford)
Nando de Freitas (DeepMind)
Jascha Sohl-Dickstein (Google Brain)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors