Gradient feedback framework for joint transceiver neural network training

A method of wireless communication by a transmitting device, comprising:

• receiving a transmission reference point gradient of a loss from a receiving device;
• determining a transmission point-payload gradient of a transmission reference point value with respect to an encoded value generated by a transmitter neural network;
• determining a payload gradient of the loss based on a product of the transmission reference point gradient and the transmission point-payload gradient;
• and
• updating weights of the transmitter neural network based on the payload gradient.

An apparatus for wireless communications performed by a receiving device, comprising:

• a processor,
• memory coupled with the processor;
• and
• instructions stored in the memory and operable, when executed by the processor, to cause the apparatus:
• to determine a transmission reference point value;
• to determine a transmission reference point gradient of a loss based on the transmission reference point value;
• and
• to transmit a message comprising the transmission reference point gradient to a transmitting device.

An apparatus for wireless communications by a transmitting device, comprising:

• a processor,
• memory coupled with the processor;
• and
• instructions stored in the memory and operable, when executed by the processor, to cause the apparatus:
• to receive a transmission reference point gradient of a loss from a receiving device;
• to determine a transmission point-payload gradient of a transmission reference point value with respect to an encoded value generated by a transmitter neural network;
• to determine a payload gradient of the loss based on a product of the transmission reference point gradient and the transmission point-payload gradient;
• and
• to update weights of the transmitter neural network based on the payload gradient.

A receiving device for wireless communications, comprising:

• means for determining a transmission reference point value;
• means for determining a transmission reference point gradient of a loss based on the transmission reference point value;
• and
• means for transmitting a message comprising the transmission reference point gradient to a transmitting device.

A transmitting device for wireless communications, comprising:

• means for receiving a transmission reference point gradient of a loss from a receiving device;
• means for determining a transmission point-payload gradient of a transmission reference point value with respect to an encoded value generated by a transmitter neural network;
• means for determining a payload gradient of the loss based on a product of the transmission reference point gradient and the transmission point-payload gradient;
• and
• means for updating weights of the transmitter neural network based on the payload gradient.

A non-transitory computer-readable medium having program code recorded thereon, the program code executed by a receiving device and comprising:

• program code to determine a transmission reference point value;
• program code to determine a transmission reference point gradient of a loss based on the transmission reference point value;
• and
• program code to transmit a message comprising the transmission reference point gradient to a transmitting device.

A non-transitory computer-readable medium having program code recorded thereon, the program code executed by a transmitting device and comprising:

• program code to receive a transmission reference point gradient of a loss from a receiving device;
• program code to determine a transmission point-payload gradient of a transmission reference point value with respect to an encoded value generated by a transmitter neural network;
• program code to determine a payload gradient of the loss based on a product of the transmission reference point gradient and the transmission point-payload gradient;
• and
• program code to update weights of the transmitter neural network based on the payload gradient.