Eigensystem optimization in artificial neural networks

Patent metadata

Field	Value
Publication number	US-11677449-B2
Application number	US-202017062970-A
Country	US
Kind code	B2
Filing date	Oct 5, 2020
Priority date	Apr 24, 2018
Publication date	Jun 13, 2023
Grant date	Jun 13, 2023

Abstract

Systems, methods, computer program products, and devices provide for computing an eigensystem from a first data set; computing updated eigenvalues that approximate an eigensystem of at least a second data set based on the eigensystem of the first data set; and evaluating a plurality of features in each of the first and at least second data sets using a cost function. The cost function uses fewer than the total number of eigenvalues and can include a condition number. The cost function can perform a coarse approximation of the eigenvalues to de-select at least one of the data sets. This can be useful for learning and/or online processing in an artificial neural network.

First claim

The invention claimed is:

1. A method of antenna selection, comprising:
performing a partial update to a first selection of antennas in an antenna array to produce a second selection of antennas in the antenna array;
computing updated eigenvalues based on fewer than a total number of eigenvalues corresponding to the first selection of antennas, the updated eigenvalues corresponding to the second selection of antennas;
computing a second Multiple Input Multiple Output (MIMO) performance based on the updated eigenvalues; and
based on a comparison between the second MIMO performance and a first MIMO performance corresponding to the first selection of antennas, performing at least one of MIMO transmission and MIMO reception from an antenna array that comprises the first selection of antennas or the second selection of antennas.

2. The method of claim 1 , wherein performing the partial update comprises at least one of adding an antenna to the first selection, deleting an antenna from the first selection, and employing a sliding window.

3. The method of claim 1 , wherein the first selection of antennas comprises at least one of a Massive MIMO antenna array, a distributed antenna system, a Cooperative MIMO array, and at least one relay.

4. The method of claim 1 , further comprising computing up to the total number of eigenvalues corresponding to the first selection.

5. The method of claim 1 , wherein the updated eigenvalues approximate an eigensystem of a modified Hermitian matrix corresponding to the second selection based on knowledge of an eigensystem of a Hermitian matrix corresponding to the first selection.

6. The method of claim 1 , wherein at least one of singular value decomposition and eigen-decomposition is used to compute the eigenvalues and the updated eigenvalues.

7. The method of claim 1 , wherein the eigenvalues are calculated from at least one of a data matrix, a channel matrix, and a covariance matrix.

8. The method of claim 1 , wherein the updated eigenvalues comprise a minimum eigenvalue and a maximum eigenvalue of the total number of eigenvalues corresponding to the second selection.

9. The method of claim 1 , wherein the MIMO performance comprises a condition number.

10. The method of claim 1 , wherein computing updated eigenvalues is performed via at least one of parallel processing and pipelined processing.

11. The method of claim 1 , wherein computing updated eigenvalues is performed using numerical computations with a selectable accuracy.

12. The method of claim 1 , wherein computing updated eigenvalues is performed using a spectrum-slicing algorithm to isolate eigenvalues to disjoint intervals and using Newton's method to search within at least one of the disjoint intervals.

13. An apparatus, comprising at least one processor, at least one memory in electronic communication with the at least one processor, and instructions stored in the at least one memory, the instructions executable by the at least one processor for:
performing a partial update to a first selection of antennas in an antenna array to produce a second selection of antennas in the antenna array;
computing updated eigenvalues based on fewer than a total number of eigenvalues corresponding to the first selection, the updated eigenvalues corresponding to the second selection;
computing a second Multiple Input Multiple Output (MIMO) performance based on the updated eigenvalues; and
based on a comparison between the second MIMO performance and a first MIMO performance corresponding to the first selection of antennas, performing at least one of MIMO transmission and MIMO reception from an antenna array that comprises the first selection of antennas or the second selection of antennas.

14. The apparatus of claim 13 , wherein performing the partial update comprises at least one of adding an antenna to the first selection, deleting an antenna from the first selection, and employing a sliding window.

15. The apparatus of claim 13 , wherein the first selection of antennas comprises at least one of a Massive MIMO antenna array, a distributed antenna system, a Cooperative MIMO array, and at least one relay.

16. The apparatus of claim 13 , further comprising instructions executable by the at least one processor for computing up to the total number of eigenvalues corresponding to the first selection.

17. The apparatus of claim 13 , wherein the updated eigenvalues approximate an eigensystem of a modified Hermitian matrix corresponding to the second selection based on knowledge of an eigensystem of a Hermitian matrix corresponding to the first selection.

18. The apparatus of claim 13 , wherein at least one of singular value decomposition and eigen-decomposition is used to compute the eigenvalues and the updated eigenvalues.

19. The apparatus of claim 13 , wherein the eigenvalues are calculated from at least one of a data matrix, a channel matrix, and a covariance matrix.

20. The apparatus of claim 13 , wherein the updated eigenvalues comprise a minimum eigenvalue and a maximum eigenvalue of the total number of eigenvalues corresponding to the second selection.

21. The apparatus of claim 13 , wherein the MIMO performance comprises a condition number.

22. The apparatus of claim 13 , wherein computing updated eigenvalues is performed via at least one of parallel processing and pipelined processing.

23. The apparatus of claim 13 , wherein computing updated eigenvalues is performed using numerical computations with a selectable accuracy.

24. The apparatus of claim 13 , wherein computing updated eigenvalues is performed using a spectrum-slicing algorithm to isolate eigenvalues to disjoint intervals and using Newton's method to search within at least one of the disjoint intervals.

Assignees

• Genghiscomm Holdings Llc
• Tybalt Llc

Inventors

• Shattil Steve

Classifications

• H04B7/0473

  taking constraints in layer or codeword to antenna mapping into account · CPC title

• H04B7/0452 Primary

  Multi-user MIMO systems · CPC title

• H04B7/0413 Primary

  MIMO systems · CPC title

• H04B7/0691

  using subgroups of transmit antennas · CPC title

• H04B7/026

  Co-operative diversity, e.g. using fixed or mobile stations as relays · CPC title