Hardware Aware Algorithm for Convolution Representation

The network architecture features multiple convolutional and fully connected layers to let use dynamics of state space models equations, simply XY+Z. This is specifically programmed in CUDA to analyze parallel workloads with various optimizations.

Implementation

Has the below beautiful abstract visualisations that compares both CPU and GPU artifacts

Want to know the implementation?

Analysis.pdf is all yours. You can observe the explanation of the implementation. Furthermore head to src/cudaLib.cu for implementation of Convolution, Pooling and Softmax layers. You can run the application by

1. Having a newer version of cuda installed.

2. cd ~/Hardware-aware-algorithm

3. cmake .

4. make

5. ./run

References

1. Alex Krizhevsky, Ilya Sutskever, and Geoffrey E. Hinton. 2012. ImageNet classification with deep convolutional neural networks. In Proceedings of the 25th International Conference on Neural Information Processing Systems - Volume 1 (NIPS'12). Curran Associates Inc., Red Hook, NY, USA, 1097–1105.

2. https://docs.nvidia.com/cuda/cuda-c-programming-guide/