Skip to content

Programming model/paradigm

Symbolic VS imperative

./Symbolic-and-imperative中进行了总结。

TensorFlow Programming Model

TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems中对TensorFlow的programming model进行了描述。

经验总结

Rule one: separation of model and dataset

将model和dataset分隔开来

Use trainer to connect dataset and model

Feedforward and feedback

前馈过程:计算得到loss

反馈过程:backpropagation to compute gradient and then adjust parameter;