Comparing floating-point and logarithmic number representations for reconfigurable acceleration | IEEE Conference Publication | IEEE Xplore