Abstract: A method for convolution calculation in a neural network is provided. The method comprises: decomposing each weight into multiple sub-weights, each with only one valid bit, representing different bit significance (bit plane); accumulating input feature map units corresponding to each of the sub-weights with the same bit significance to obtain intermediate sums; shifting each of the intermediate sums according to the bit significance of the corresponding sub-weights to obtain shifted intermediate sums; and accumulating the shifted intermediate sums.

Abstract: A system for offset cancellation for driving a display panel includes: a plurality of source amplifiers driving the display panel; an image analyzer configured to receive a data input of an image frame and analyze the data input; and a chopping pattern controller connected with the image analyzer and configured to determine a chopping pattern that fits the data input based on analysis results of the image analyzer, and apply the determined chopping pattern to the source amplifiers. The source amplifiers are divided into N groups while the chopping pattern controller is configured to drive source amplifiers in each group by a single chopping control signal. The image analyzer is configured to generate an indicator that indicates whether image data being analyzed corresponds to a general image or one of pre-registered killer pattern images. A method for offset cancellation for driving a display panel is also provided.