METHOD, SYSTEM, AND PROGRAM FOR CONVERTING APPLICATION PROGRAM CODE TO EXECUTABLE CODE USING NEURAL NETWORKS BASED ON CHARACTERISTICS OF THE INPUTS

• US 20040122785A1
• Filed: 12/15/2000
• Published: 06/24/2004
• Est. Priority Date: 12/15/2000
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Nguyen, Chung Tien, Brown, Michael Wayne
• Granted Patent: US 6,826,550 B2
• Time in Patent Office: Days
• US Class Current: 706/16
• CPC Class Codes: G06F 8/427 Parsing
• CPC Class Codes: G06F 8/52 Binary to binary
• CPC Class Codes: G06N 3/105 Shells for specifying net layout

Abstract

Provided is a compiler to map application program code to object code capable of being executed on an operating system platform. A first neural network module is trained to generate characteristic output based on input information describing attributes of the application program. A second neural network module is rained to receive as input the application program code and the characteristic output and, in response, generate object code. The first and second neural network modules are used to convert the application program code to object code.

Claims

1. A method for generating a compiler to map application program code to object code capable of being executed on an operating system platform, comprising:

training a first neural network module to generate characteristic output based on input information describing attributes of the application program;

training a second neural network module to receive as input the application program code and the characteristic output and, in response, generate object code; and

using the first and second neural network modules to convert the application program code to object code.

2. The method of claim 1, wherein the application program code comprises a Java Applet bytecode implementing a Java Applet class.

3. The method of claim 1, wherein the second neural network module comprises a first and second neural network components, wherein training the second neural network module further comprises:

training the first neural network component to map the application program code to a unique representation of the application program code; and

training the second neural network component to generate executable code from the unique representation of the application program code and the characteristic output.

4. The method of claim 3, wherein the first neural network module comprises a plurality of component neural networks, wherein training the component neural networks of the first module comprises:

parsing the application program code into a plurality of input vectors; and

assigning each input vector to one of the component neural networks of the first module, wherein each of the component neural networks of the first module are trained to convert the assigned input vector to an output vector, wherein the combination of the output vectors from each component neural network forms a unique representation of the application code.

5. The method of claim 4, wherein the application program code comprises a Java Applet bytecode implementing a Java Applet class, and wherein parsing the application program into a plurality of input vectors, comprises parsing the application program code into categories based on the bytecode class file format, wherein each input vector includes bits from the bytecode that comprise one of the categories.

6. The method of claim 1, wherein the first neural network module comprises a plurality of characterization neural networks, and wherein training the characterization neural networks comprises:

determining attribute information of the application program code;

transmitting the attribute information to each characterization neural network to determine an application type to which the application program is capable of belonging, wherein the characteristic output indicates the application type for the application program.

7. The method of claim 6, wherein the application types to which the application program is capable of belonging includes at least one of a business type, scientific type, audio type, graphics type, and security type.

8. The method of claim 6, wherein the attribute information for the application program includes at least one of a request time, security level, the operating system platform type, descriptive terms describing aspects of the application program, and the system user profile.

9. The method of claim 8, wherein the operating system platform is implemented in a client computer that downloads the application program from a server over a network to execute, and wherein the request time comprises the time the client computer requested the application program from the server.

10. The method of claim 8, wherein the operating system platform is implemented in a computer that accesses the application program from a local storage device to execute, and wherein the request time comprises the time the computer executed the application program.

11. The method of claim 6, wherein the characteristic output indicates a likelihood that the application program is each of a set of application types.

12. The method of claim 3, wherein the second neural network component comprises a recurrent neural network, and wherein training the second neural network component comprises performing iterations inputting the unique representation of the application program code and the characteristic output to the second neural network component to generate the object code, wherein the generated object code is fed back into the second neural network component until the second neural network component is capable of mapping the unique representation of the code set to the object code with a minimal amount of error in the mapping.

13. A method for mapping application program code to object code capable of being executed on an operating system platform, comprising:

receiving the application program code with a first neural network module;

generating, with the first neural network module, characteristic output based on input information describing attributes of the application program;

receiving, with a second neural network module, the application program code and the characteristic output and generating therefrom the object code.

14. The method of claim 13, wherein the second neural network module comprises a first and second neural network components, further comprising:

mapping, with the first neural network component, the application program code to a unique representation of the application program code; and

generating, with the second neural network component, an executable code from the unique representation of the application program code and the characteristic output.

16. The method of claim 14, wherein the first neural network module comprises a plurality of characterization neural networks, and wherein generating the characteristic output comprises:

determining attribute information of the application program code;

transmitting the attribute information to each characterization neural network to determine an application type to which the application program belongs.

15. The method of claim 15, wherein the second neural network component comprises a plurality of neural networks, wherein converting the application program code to the unique representation with the second neural network component comprises:

parsing the application program code into a plurality of input vectors;

assigning each input vector to one of the neural networks of the second neural network component;

converting, with each of the neural networks in the second neural network component, the assigned input vector to an output vector; and

combining the output vectors to form a unique representation of the code set.

17. A system for mapping application program code to object code capable of being executed on an operating system platform, comprising:

(a) a first neural network module including logic that performs;

(i) receiving the application program code with

(ii) generating characteristic output based on input information describing attributes of the application program;

(b) a second neural network module including logic that performs;

(i) receiving the application program code and the characteristic output; and

(ii) generating from the received application program code and the characteristic output the object code.

18. The system of claim 17, wherein the second neural network module comprises:

a first neural network component including logic mapping the application program code to a unique representation of the application program code; and

a second neural network component including the logic generating the object code from the unique representation of the application program code and the characteristic output.

19. The system of claim 18, wherein the second neural network component further comprises:

a plurality of neural networks; and

wherein the logic for converting the application program code to the unique representation performs;

(i) parsing the application program code into a plurality of input vectors;

(ii) assigning each input vector to one of the neural networks of the second neural network component;

(iii) converting, with each of the neural networks in the second neural network component, the assigned input vector to an output vector; and

(iv) combining the output vectors to form a unique representation of the code set.

20. The system of claim 18, wherein the first neural network module further comprises:

a plurality of characterization neural networks; and

wherein the logic for generating the characteristic output comprises;

(i) determining attribute information of the application program code; and

(ii) transmitting the attribute information to each characterization neural network to determine an application type to which the application program belongs.

21. The system of claim 20, wherein the application types to which the application program is capable of belonging includes at least one of a business type, scientific type, audio type, graphics type, and security type.

22. The system of claim 20, wherein the attribute information for the application program includes at least one of a request time, security level, the operating system platform type, descriptive terms describing aspects of the application program, and the system user profile.

23. The system of claim 20, wherein the characteristic output indicates a likelihood that the application program is each of a set of application types.

24. The system of claim 17, wherein the neural network modules are implemented in a software program.

25. The system of claim 24, wherein the software program is implemented in a Java Just-In-Time Compiler component of a Java Virtual Machine.

26. The system of claim 17, wherein the neural network modules are implemented in at least one integrated circuit device.

27. An article of manufacture for generating a compiler to map application program code to object code capable of being executed on an operating system platform, wherein the article of manufacture includes training code capable of causing a processor to perform:

training a first neural network module to generate characteristic output based on input information describing attributes of the application program;

training a second neural network module to receive as input the application program code and the characteristic output and, in response, generate object code.

28. The article of manufacture of claim 27, wherein the second neural network module comprises a first and second neural network components, and wherein the training code for training the second neural network module further comprises:

training the first neural network component to map the application program code to a unique representation of the application program code; and

training the second neural network component to generate executable code from the unique representation of the application program code and the characteristic output.

29. The article of manufacture of claim 27, wherein the first neural network module comprises a plurality of component neural networks, wherein the training code for training the first neural network module further performs:

parsing the application program code into a plurality of input vectors; and

assigning each input vector to one of the component neural networks, wherein each of the component neural networks are trained to convert the assigned input vector to an output vector, wherein the combination of the output vectors from each component neural network forms a unique representation of the application code.

30. The article of manufacture of claim 29, wherein the application program code comprises a Java Applet bytecode implementing a Java Applet class, and wherein parsing the application program into a plurality of input vectors, comprises parsing the application program code into categories based on the bytecode class file format, wherein each input vector includes bits from the bytecode that comprise one of the categories.

31. The article of manufacture of claim 27, wherein the first neural network module comprises a plurality of characterization neural networks, and wherein training the first neural network module further comprises:

determining attribute information of the application program code;

transmitting the attribute information to each characterization neural network to determine an application type to which the application program is capable of belonging, wherein the characteristic output indicates the application type for the application program.

32. The article of manufacture of claim 31, wherein the application types to which the application program is capable of belonging includes at least one of a business type, scientific type, audio type, graphics type, and security type.

33. The article of manufacture of claim 31, wherein the attribute information for the application program includes at least one of a request time, security level, the operating system platform type, descriptive terms describing aspects of the application program, and the system user profile.

34. The article of manufacture of claim 31, wherein the characteristic output indicates a likelihood that the application program is each of a set of application types.

35. The article of manufacture of claim 28, wherein the second neural network component comprises a recurrent neural network, and wherein training the second neural network component comprises performing iterations inputting the unique representation of the application program code and the characteristic output to the second neural network component to generate the object code, wherein the generated object code is fed back into the second neural network component until the second neural network component is capable of mapping the unique representation of the code set to the object code with a minimal amount of error in the mapping.