PURPOSE: To adjust weight in real time by providing first and second floating gate devices at a semiconductor circuit.
CONSTITUTION: Each synapse cell includes a pair of floating gate devices 20 and 21 which function as a learning circuit for the parallel realization of a learning algorithm. For example, a synapse cell 25a includes floating gate devices 20a and 21b. Also, each floating gate is connected between a summary line S1 and a summary line S2, and one common floating gate member is shared. Moreover, the control gate of the floating device 20a is connected with a programming input word line 17, and the control gate of the floating gate device 21a is connected with a programming input word line 18. Then, a method for realizing weight adjustment in parallel uses a device technique for executing weight update in parallel in a prescribed number of cycles.