Accelerating small matrix multiplications by adaptive batching strategy on GPU | IEEE Conference Publication | IEEE Xplore