Since this paper is similar to paper ID 6113, this weakens the novelty of the paper as set out in the dual submission policy. However, the paper has more novel contributions over 6113, such as using k convolutions instead of 1 convolution. The paper would also benefit from a more complete description of the main method.