作者: James Bergstra , Olivier Breuleux , Frédéric Bastien , Pascal Lamblin , Razvan Pascanu
DOI: 10.25080/MAJORA-92BF1922-003
关键词:
摘要: Theano is a compiler for mathematical expressions in Python that combines the convenience of NumPy's syntax with speed optimized native machine language. The user composes high-level description mimics and semantics, while being statically typed functional (as opposed to imperative). These allow provide symbolic differentiation. Before performing computation, optimizes choice expressions, translates them into C++ (or CUDA GPU), compiles dynamically loaded modules, all automatically. Common learn- ing algorithms implemented are from 1:6 7:5 faster than competitive alternatives (including those C/C++, NumPy/SciPy MATLAB) when compiled CPU between 6:5 44 GPU. This paper illustrates how use Theano, outlines scope compiler, provides benchmarks on both GPU processors, explains its overall design.