Improving Forward Gradient Learning for Benchmarking Deep Neural Networks - 42Papers