SYSTEM AND METHOD FOR INCREASING ACCURACY OF APPROXIMATING QUERY RESULTS USING NEURAL NETWORKS

• US 20190050454A1
• Filed: 12/29/2017
• Published: 02/14/2019
• Est. Priority Date: 08/14/2017
• Status: Active Grant

• Current Assignee: Sisense Ltd. (Sisense Inc.)
• Original Assignee: Sisense Ltd. (Sisense Inc.)
• Inventors: ORAD, Amir, REGEV, Nir, LEVY YURISTA, Guy, AZARIA, Adi
• Granted Patent: US 10,642,835 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/24549 Run-time optimisation
• CPC Class Codes: G06F 16/2455 Query execution
• CPC Class Codes: G06F 16/2462 Approximate or statistical queries
• CPC Class Codes: G06F 16/248 Presentation of query results
• CPC Class Codes: G06F 16/903 Querying for retrieval from the web G06F16/953
• CPC Class Codes: G06F 16/90335 Query processing
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/06 Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent

Abstract

A system and method for increasing the accuracy of generated approximations of query results. The method includes sending a received query to a first neural network and a second neural network; receiving from the first neural network a first predicted result to the received query; providing the first predicted result as a first output to a device having initiated the received query; receiving from the second neural network a second predicted result to the received query; and providing the second predicted result as a second output to the device having initiated the received query, wherein the first neural network requires less computational resources than the second neural network, and whereby the first output is provided before the second output.

Claims

1. A method for improving accuracy of generated approximations of query results, comprising:

sending a received query to a first neural network and a second neural network;

receiving from the first neural network a first predicted result to the received query;

providing the first predicted result as a first output to a device having initiated the received query;

receiving from the second neural network a second predicted result to the received query; and

providing the second predicted result as a second output to the device having initiated the received query, wherein the first neural network requires less computational resources than the second neural network, and whereby the first output is provided before the second output.

2. The method of claim 1, further comprising:

querying the database with the received query to retrieve a real result.

3. The method of claim 2, wherein the second predicted result has a margin of error lower than a margin of error of the first predicted result.

4. The method of claim 2, further comprising:

determining if the second predicted result exceeds a predetermined threshold of the real result.

5. The method of claim 4, further comprising:

providing the real result as a third output, when the second predicted result exceeds the predetermined threshold of the real result.

6. The method of claim 4, further comprising:

providing the second output when the second output exceeds the predetermined threshold.

7. The method of claim 1, wherein the first neural network and the second neural network each include a plurality of neurons configured to generate the first predicted result and the second predicted result respectively, and wherein each neuron is assigned a weight value.

8. The method of claim 7, wherein the second neural network includes more neurons than the first neural network.

9. The method of claim 1, wherein the first predicted result and the second predicted result are provided without querying the database with the received query.

10. The method of claim 1, further comprising:

determining a confidence level for the first predicted result; and

sending the received query to the second neural network only when the confidence level is below a first predefined threshold.

11. The method of claim 10, wherein the confidence level is determined using a loss function.

12. The method of claim 1, further comprising:

receiving the query and providing any of the first output, the second output, and the third output to any one of;

a user node and an approximation server.

13. The method of claim 12, wherein the approximation server is connected to a plurality of user devices and a plurality of neural networks.

14. A non-transitory computer readable medium having stored thereon instructions for causing a processing circuitry to perform a process, the process comprising:

sending a received query to a first neural network and a second neural network;

receiving from the first neural network a first predicted result to the received query;

providing the first predicted result as a first output to a device having initiated the received query;

receiving from the second neural network a second predicted result to the received query; and

providing the second predicted result as a second output to the device having initiated the received query, wherein the first neural network requires less computational resources than the second neural network, and whereby the first output is provided before the second output.

15. A system for increasing the accuracy of generated approximations of query results, comprising:

a processing circuitry; and

a memory, the memory containing instructions that, when executed by the processing circuitry, configure the system to;

send a received query to a first neural network and a second neural network;

receive from the first neural network a first predicted result to the received query;

provide the first predicted result as a first output to a device having initiated the received query;

receive from the second neural network a second predicted result to the received query; and

provide the second predicted result as a second output to the device having initiated the received query, wherein the first neural network requires less computational resources than the second neural network, and whereby the first output is provided before the second output.

16. The system of claim 15, wherein the system is further configured to:

query the database with the received query to retrieve a real result.

17. The system of claim 16, wherein the second predicted result has a margin of error lower than a margin of error of the first predicted result.

18. The system of claim 16, wherein the system is further configured to:

determine if the second predicted result exceeds a predetermined threshold of the real result.

19. The system of claim 18, wherein the system is further configured to:

provide the real result as a third output, when the second predicted result exceeds the predetermined threshold of the real result.

20. The system of claim 18, wherein the system is further configured to:

provide the second output when the second output exceeds the predetermined threshold.

21. The system of claim 15, wherein the first neural network and the second neural network each include a plurality of neurons configured to generate the first predicted result and the second predicted result respectively, and wherein each neuron is assigned a weight value.

22. The system of claim 21, wherein the second neural network includes more neurons than the first neural network.

23. The system of claim 15, wherein the first predicted result and the second predicted result are provided without querying the database with the received query.

24. The system of claim 15, wherein the system is further configured to:

determine a confidence level for the first predicted result; and

send the received query to the second neural network only when the confidence level is below a first predefined threshold.

25. The system of claim 24, wherein the confidence level is determined using a loss function.

26. The system of claim 15, wherein the system is further configured to:

receive the query and providing any of the first output, the second output, and the third output to any one of;

a user node and an approximation server.

27. The system of claim 26, wherein the approximation server is connected to a plurality of user devices and a plurality of neural networks.