Systems and methods of tuning sensor devices
Automated methods and systems for tuning parameters of detectors based on metrics are provided. The tuning methods and systems include using machine learning to adjust registers used for setting detector parameters in an automated fashion to achieve optimal performance in a time-efficient manner. The tuning process can include determining available registers provided by the detector and setting values within those registers to a starting value. A binary search in either direction from the starting value is then performed to determine the limits of the parameter values for each register. The register values can then be optimized in view of selected metrics.
1. A system comprising:
a processor; and
a computer-readable storage medium storing computer-readable instructions which, when executed by the processor, cause the processor to execute a method, the method comprising:
identifying a respective starting value for each of a plurality of registers, wherein each register stores a value of a parameter of a detector system;
determining, for each register, a respective effective range of reasonable returns by binary searching to find an upper limit of reasonable returns and a lower limit of reasonable returns;
determining, for each register, a respective optimal value within the respective effective range of reasonable returns by optimizing one or more metrics over the respective effective range of reasonable returns for each register; and
setting each register to the respective optimal value.
2. The system of claim 1 , wherein the respective starting value is manually set.
3. The system of claim 1 , wherein determining the respective effective range of reasonable returns comprises, for each register:
leaving all registers in a starting configuration;
binary searching upward to find the upper limit of reasonable returns;
binary searching downward to find the lower limit of reasonable returns;
saving a result of the binary searching upward and the binary searching downward as the respective effective range of reasonable returns; and
returning to the starting configuration.
4. The system of claim 1 , wherein optimizing the one or more metrics over the respective effective range of reasonable returns comprises, for a first set of registers associated with a first set of metrics, optimizing metrics of the first set of metrics, and setting each register of the first set of registers to the respective optimal value.
5. The system of claim 1 , wherein setting each register to the respective optimal value comprises:
initializing each register to a value within the respective effective range of reasonable returns;
optimizing metrics over the respective effective range for each register; and
setting each register to the optimized metrics.
6. The system of claim 1 , wherein the method further comprises using the respective optimal value of each register to tune a detector.
7. The system of claim 1 , wherein the respective optimal value within the respective effective range for each register is determined based on one or more of a gradient descent, a direct search, a Latin hypercube algorithm, and a Bayesian optimization algorithm.
8. A method for tuning a detector, the method comprising:
identifying a respective starting value for each of a plurality of registers;
determining, for each register, a respective effective range of reasonable returns by binary searching to find an upper limit of reasonable returns and a lower limit of reasonable returns;
determining, for each register, a respective optimal value within the respective effective range of reasonable returns by optimizing one or more metrics over the respective effective range of reasonable returns for each register; and
setting each register to the respective optimal value.
9. The method of claim 8 , wherein the respective starting value is manually set.
10. The method of claim 8 , wherein determining the respective effective range of reasonable returns comprises, for each register:
leaving all registers in a starting configuration;
binary searching upward to find the upper limit of reasonable returns;
binary searching downward to find the lower limit of reasonable returns;
saving a result of the binary searching upward and the binary searching downward as the respective effective range of reasonable returns; and
returning to the starting configuration.
11. The method of claim 8 , wherein optimizing the one or more metrics over the respective effective range of reasonable returns comprises, for a first set of registers associated with a first set of metrics, optimizing metrics of the first set of metrics, and setting each register of the first set of registers to the respective optimal value.
12. The method of claim 8 , wherein setting each register to the respective optimal value comprises:
initializing each register to a value within the respective effective range of reasonable returns;
optimizing metrics over respective effective range for each register; and
setting each register to the optimized metrics.
13. The method of claim 8 , further comprising using the respective optimal value of each register to tune a detector.
14. The method of claim 8 , wherein the respective optimal value within the respective effective range for each register is determined based on one or more of a gradient descent, a direct search, a Latin hypercube algorithm, and a Bayesian optimization algorithm.
15. A tuning system for a sensor, the system comprising:
at least one sensor;
a processor; and
a computer-readable storage medium storing computer-readable instructions which, when executed by the processor, cause the processor to execute a method, the method comprising:
identifying a respective starting value for each of a plurality of registers, wherein each register stores a value of a parameter of a detector system;
determining, for each register, a respective effective range of reasonable returns by binary searching to find an upper limit of reasonable returns and a lower limit of reasonable returns;
determining, for each register, a respective optimal value within the respective effective range of reasonable returns by optimizing one or more metrics over the respective effective range of reasonable returns for each register; and
setting each register to respective optimal value.
16. The system of claim 15 , wherein the respective starting value is manually set.
17. The system of claim 15 , wherein determining the respective effective range of reasonable returns comprises, for each register:
leaving all registers in a starting configuration;
binary searching upward to find the upper limit of reasonable returns;
binary searching downward to find the lower limit of reasonable returns;
saving a result of the binary searching upward and the binary searching downward as the respective effective range of reasonable returns; and
returning to the starting configuration.
18. The system of claim 15 , wherein optimizing the one or more metrics over the respective effective range of reasonable returns comprises, for a first set of registers associated with a first set of metrics, optimizing metrics of the first set of metrics, and setting each register of the first set of registers to the respective optimized value.
19. The system of claim 15 , wherein setting each register to the respective optimal value comprises:
initializing each register to a value within the respective effective range of reasonable returns;
optimizing metrics over the respective effective range for each register; and
setting each register to the optimized metrics.
20. The system of claim 15 , wherein the method further comprises using the respective register values to tune a detector.