BriefGPT.xyz
Sep, 2017
FiLM:通用调节层的视觉推理
FiLM: Visual Reasoning with a General Conditioning Layer
HTML
PDF
Ethan Perez, Florian Strub, Harm de Vries, Vincent Dumoulin, Aaron Courville
TL;DR
引入了一种名为FiLM的神经网络通用条件方法,FiLM层通过基于条件信息的简单,特征-wise仿射变换影响神经网络计算,该方法对于视觉推理任务特别有效,在CLEVR基准测试中减少了一半的错误率,FiLM层总体上能够很好的适应少样例情况下的新数据以及零样例的情况。
Abstract
We introduce a general-purpose conditioning method for
neural networks
called
film
: Feature-wise Linear Modulation.
film
layers influence
→