Chainer  FAQ.docx

Version 1

    Chainer is a flexible framework for neural networks. It uses a dynamic computational graph that uses the define by run scheme, instead of the traditional define and run scheme. In the define by run approach, the network is defined dynamically via the actual forward computation. It stores the
    computation history instead of programming logic, which enables us to fully leverage the power of programming logic in Python*. More information can be
    found in the Define-by-Run guide.