Method and apparatus for adapting feature data in a convolutional neural network

An apparatus for adapting feature data in a convolutional neural network, comprising:

• a selector configured to select a plurality of consecutive layers in the convolutional neural network, a sum of sizes of output feature data and related parameters of a last layer in the plurality of consecutive layers being less than or equal to a capacity of a cache for caching data relating to operations of the convolutional neural network;
• and
• a splitter configured to
• determine an expected number of subdata blocks and a layout position, a width and a height of each subdata block of the output feature data of the last layer;
• determine, for each current layer in the plurality of layers starting from the last layer, a layout position, a width, and a height of each subdata block of an input feature data for the current layer according to the layout position, the width, and the height of each subdata block of the output feature data of the current layer, until the layout position, the width, and the height of each subdata block of the input feature data for a first layer in the plurality of layers are determined, each subdata block of the input feature data for the first layer being capable of entirely stored in the cache;
• and
• determine an actual position of each subdata block of the input feature data for the first layer in the input feature data of the first layer.