Could an artificial neural network algorithm be expressed in terms of map-reduce
ID: 649128 • Letter: C
Question
Could an artificial neural network algorithm be expressed in terms of map-reduce operations? I am also interested more generally in methods of parallelization as applied to ANNs and their application to cloud computing.
I would think one approach would involve running a full ANN on each node and somehow integrating the results in order to treat the grid like a single entity (in terms of input/output and machine learning characteristics.) I would be curious even in this case what such an integrating strategy might look like
Explanation / Answer
Yes it can, and has been. In the paper Map-Reduce for Machine Learning on Multicore they discuss using the Map-Reduce paradigm for several common ML algorithms including ANNs.