This software is a Python-based optimizing compiler that speeds up the evaluation of complex expressions.
I recently had the chance to try out Theano, an open source Python based tool that promises to allow you to define, optimize, and evaluate mathematical expressions with ease. After using it for a while, I have been extremely impressed by its performance and versatility.
One of the things that really stood out to me was the speed improvements Theano offers over using pure NumPy. In fact, it is not uncommon to see speed improvements of ten-fold or more! I attribute this to Theano's ability to meld some aspects of a computer algebra system (CAS) with aspects of an optimizing compiler. This combination of CAS with optimizing compilation is particularly useful for computational fields in which complicated mathematical expressions are evaluated repeatedly, and evaluation speed is critical.
Theano supports a range of numerical types in multiple dimensions and a number of well-tested operations, making it a great tool for complex computational tasks involving multi-dimensional arrays. One particularly useful feature is the ability to compute the gradient of an expression with respect to another.
Symbolic expressions can be compiled into functions that work on the same data structures as NumPy, allowing for easy interoperability. Theano's compiler applies many optimizations of varying complexity to these symbolic expressions. Some of these optimizations include constant folding, merging of similar subgraphs to avoid unnecessary calculations, arithmetic simplification, inserting efficient BLAS operations, and using inplace operations wherever possible.
Overall, I would highly recommend Theano to anyone looking for a powerful and efficient tool for mathematical computation. Its impressive speed and flexibility make it a great choice for a wide range of applications, from machine learning to data analysis and scientific research.
One of the things that really stood out to me was the speed improvements Theano offers over using pure NumPy. In fact, it is not uncommon to see speed improvements of ten-fold or more! I attribute this to Theano's ability to meld some aspects of a computer algebra system (CAS) with aspects of an optimizing compiler. This combination of CAS with optimizing compilation is particularly useful for computational fields in which complicated mathematical expressions are evaluated repeatedly, and evaluation speed is critical.
Theano supports a range of numerical types in multiple dimensions and a number of well-tested operations, making it a great tool for complex computational tasks involving multi-dimensional arrays. One particularly useful feature is the ability to compute the gradient of an expression with respect to another.
Symbolic expressions can be compiled into functions that work on the same data structures as NumPy, allowing for easy interoperability. Theano's compiler applies many optimizations of varying complexity to these symbolic expressions. Some of these optimizations include constant folding, merging of similar subgraphs to avoid unnecessary calculations, arithmetic simplification, inserting efficient BLAS operations, and using inplace operations wherever possible.
Overall, I would highly recommend Theano to anyone looking for a powerful and efficient tool for mathematical computation. Its impressive speed and flexibility make it a great choice for a wide range of applications, from machine learning to data analysis and scientific research.
What's New
Version 0.1: N/A