Method for neural network and apparatus performing same method

A method which is performable by using a neural network, the method comprising:

• determining a current activation of an artificial neuron based on a previous activation of the artificial neuron and an input signal received via an input synapse of the artificial neuron;
• determining whether an event occurs in response to a first cluster that includes the previous activation and a second cluster that includes the determined current activation;
• and
• transmitting, to an output synapse, an output signal that corresponds to the event in response to an occurrence of the event.

A method which is performable by using a recurrent neural network, the method comprising:

• obtaining a first input delta vector that corresponds to a first time based on a difference between a second input vector that corresponds to a second time that is earlier than the first time and a first input vector the corresponds to the first time;
• obtaining a hidden state delta vector at the corresponds to the second time based on a difference between a third hidden state vector that corresponds to a third time that is earlier than the second time and a second hidden state vector that corresponds to the second time;
• determining at least one parameter of the recurrent neural network based on a multiplication of a first weight that corresponds to the first input vector and a second weight that corresponds to the obtained first input delta vector, and a multiplication of a third weight that corresponds to the second hidden state vector and a fourth weight that correspond to the obtained hidden state delta vector;
• and
• determining a first hidden state vector that corresponds to the first time based on the determined at least one parameter of the recurrent neural network.

An electronic device comprising:

• a plurality of processing components that respectively correspond to artificial neurons,
• wherein each of the processing components includes a respective processor and a respective memory configured to store respective computer-readable instructions, and
• wherein for each respective processor, when the corresponding instructions are executed by the respective processor, the respective processor is configured to:
• determine a current activation based on a previous activation and an input signal received via an input link,
• determine a change amount of a first activation based on the current activation and a second activation that corresponds to an event that previously occurred,
• determine whether a new event occurs based on a first threshold and the determined change amount of the first activation, and
• transmit, to an output link, an output signal that corresponds to the new event in response to an occurrence of the new event.