Transmission filtering using machine learning
Systems and methods for transmission filtering are provided. A receiver includes an input coupled to a transmission line to receive distorted optical symbols. A distortion filter is coupled to the input to replace the distorted optical symbols with predicted symbols using a trained neural network. A decoder is coupled to the distortion filter to decode the predicted symbols.
1. A receiver, comprising:
an input coupled to a transmission line to receive distorted optical symbols, the distorted optical symbols being distorted using a plurality of bandwidth-limited filters configured to distort transmitted transmitter signals;
a distortion filter coupled to the input to replace the distorted optical symbols with predicted symbols using a trained neural network, the neural network being trained for functions of filter narrowing induced inter-symbol interference (FN-ISI) using only transmitter symbols and receiver symbols independently of system information; and
a decoder coupled to the distortion filter configured to decode the predicted symbols.
2. The receiver as recited in claim 1 , wherein the distortion filter compensates for filter narrowing effects.
3. The receiver as recited in claim 1 , wherein the distortion filter predicts filter narrowing induced inter-symbol interference (FN-ISI).
4. The receiver as recited in claim 1 , wherein the trained neural network is trained independently of system information, where system information includes one or more of:
baud rate, number of filters, filter bandwidth, and/or optical signal to noise ratio (OSNR).
5. The receiver as recited in claim 1 , further comprising a feedback sensor to measure performance of the predicted symbols.
6. The receiver as recited in claim 5 , further comprising a training mode triggered by the feedback sensor if performance of the predicted symbols drops below a threshold.
7. A method for transmission filtering in a receiver, comprising:
receiving distorted optical symbols over a transmission line, the distorted optical symbols being distorted using a plurality of bandwidth-limited filters configured to distort transmitted transmitter signals;
identifying the distorted symbols using a trained neural network, which outputs predicted symbols, the neural network being trained for functions of filter narrowing induced inter-symbol interference (FN-ISI) using only transmitter symbols and receiver symbols independently of system information; and
preventing filter narrowing by filtering the distorted symbols by decoding the predicted symbols in place of the distorted symbols.
8. The method as recited in claim 7 , wherein filtering the distorted symbols compensates for filter narrowing effects.
9. The method as recited in claim 7 , wherein filtering the distorted symbols predicts filter narrowing induced inter-symbol interference (FN-ISI).
10. The method as recited in claim 7 , wherein the transmission line includes an optical transmission line.
11. The method as recited in claim 7 , wherein the neural network is trained independently of system information, where system information includes one or more of: baud rate, number of filters, filter bandwidth, and/or optical signal to noise ratio (OSNR).
12. The method as recited in claim 7 , further comprising measuring performance of the predicted symbols using a feedback sensor.
13. The method as recited in claim 12 , wherein measuring performance includes measuring an error rate.
14. The method as recited in claim 12 , wherein measuring performance includes measuring a quality factor.
15. The method as recited in claim 12 , further comprising triggering a training mode by the feedback sensor if performance of the predicted symbols drops below a threshold.
16. A method for transmission filtering in a receiver, comprising:
inputting symbols received over a transmission system to a trained neural network as an input;
adjusting weights to program the neural network to yield transmitted symbols transmitted over the transmission system, the transmitted symbols being distorted into distorted optical symbols using a plurality of bandwidth-limited filters configured to distort transmitted transmitter signals; and
mitigating filter narrowing over a transmission line by:
identifying distorted symbols over the transmission line using the trained neural network, which outputs predicted symbols, the neural network being trained for functions of filter narrowing induced inter-symbol interference (FN-ISI) using only transmitter symbols and receiver symbols independently of system information; and
filtering the distorted symbols by decoding the predicted symbols in place of the distorted symbols.
17. The method as recited in claim 16 , wherein filtering narrowing includes filter narrowing induced inter-symbol interference (FN-ISI).
18. The method as recited in claim 16 , wherein the neural network is trained independently of system information, where system information includes one or more of:
baud rate, number of filters, filter bandwidth, and/or optical signal to noise ratio (OSNR).
19. The method as recited in claim 16 , further comprising measuring performance of the predicted symbols using a feedback sensor.
20. The method as recited in claim 19 , further comprising triggering a training mode by the feedback sensor if performance of the predicted symbols drops below a threshold.