GitHub

Inefficient 2D convolution compared to PyTorch

When benchmarking 2D depthwise convolutions on an NVIDIA H200, I observed that TensorFlow’s implementation is noticeably slower and consumes more power compared to PyTorch. Using a kernel-level ...
GitHub

Inefficient 2D convolution compared to JAX

When comparing a 2D depthwise convolution implemented in PyTorch vs. pure JAX/XLA on GPU, I observed that the PyTorch version runs roughly 3× slower and draws substantially more power than the ...
IEEE

Optimizing Kernel Configurations and Convolutional Strategies for Efficient Shallow CNNs in Real-Time Vision Systems

Abstract: This study investigates the impact of kernel size configurations and convolutional layer types on the performance of shallow convolutional neural networks (CNNs) for image classification ...
IEEE

Inference Efficient Source Separation Using Input-dependent Convolutions

Abstract: This paper proposes a large-capacity but inference-efficient source separation model that considers input mixtures. Convolutional neural networks (CNNs) are fundamental elements for building ...