"Convolution Neural Networks (CNN) are state of the art machine learning algorithms used in many classification applications. This class of algorithms are computationally intensive and require high memory bandwidth, making them unsuitable for embedded systems. Coarse Grain Reconfigurable Architectures (CGRAs) provides high performance and power efficient. They are capable of exploiting large amount of parallelism available in CNN algorithms. However mapping algorithms on CGRAs is a challenging task while achieving the desired performance. 
In this research we want to develop algorithms for mapping learning algorithms on CGRAs. We want mapping algorithms generic to consider the configuration parameters of the CNN and architecture parameters of CGRA"