Training machine learning by sequential conditional generalized iterative scaling

• US 7,107,207 B2
• Filed: 06/19/2002
• Issued: 09/12/2006
• Est. Priority Date: 06/19/2002
• Status: Expired due to Fees

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Goodman, Joshua Theodore
• Primary Examiner(s): Storm, Donald L.
• Application Number:
  US10/175,430
• Publication Number:
  US 20030236662A1
• Time in Patent Office: 1,546 Days
• Field of Search: None
• US Class Current: 704/9
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G10L 15/063 Training

Abstract

A system and method facilitating training machine learning systems utilizing sequential conditional generalized iterative scaling is provided. The invention includes an expected value update component that modifies an expected value based, at least in part, upon a feature function of an input vector and an output value, a sum of lambda variable and a normalization variable. The invention further includes an error calculator that calculates an error based, at least in part, upon the expected value and an observed value. The invention also includes a parameter update component that modifies a trainable parameter based, at least in part, upon the error. A variable update component that updates at least one of the sum of lambda variable and the normalization variable based, at least in part, upon the error is also provided.

Claims

1. A system for training a machine learning system, comprising:

an expected value update component that, for a plurality of outputs and for a plurality of instances in which a single feature function is non-zero, modifies an expected value based, at least in part, upon the single feature function of an input vector and an output value, a sum of lambda variable and a normalization variable;

an error calculator that calculates an error based, at least in part, upon the expected value and an observed value, the error calculation further employing, at least in part, the following equation;

2. The system of claim 1, the error calculation further employing, at least in part, the following equation:

3. The system of claim 1, modification of the expected value being based, at least in part, upon the following equation:

expected value=expected value+ƒ

_i({overscore (x)}_j, y)e^s[j,y]/z[j]

where ƒ

_i({overscore (x)}_j, y) is the feature function,

{overscore (x)}_j is the input vector,

y is the output value,

s[j,y] is the sum of lambda variable, and,

z[j] is the normalization variable.

4. The system of claim 1, the error being based, at least in part, upon the following equation:

5. The system of claim 1, modification of the trainable parameter being based, at least in part, upon the following equation:

λ

_i=λ

_i+δ

_i

where λ

_i is the trainable parameter, and,

δ

_i is the error.

6. The system of claim 1, updating of the sum of lambda variable and the normalization variable being based, at least in part, upon the following equations:

z[j]=z[j]−

e^s[j,y]

s[j,y]=s[j,y]+δ

_i

z[j]=z[j]+e^s[j,y]

where s[j,y] is the sum of lambda variable,

z[j] is the normalization variable, and,

δ

_i is the error.

7. The system of claim 1, further comprising a training data store that stores at least one of the observed value and the input vector.

8. The system of claim 7, at least one of the observed value and the input vector being stored in a sparse representation.

9. The system of claim 1, further comprising a parameter store that stores at least one trainable parameter.

10. A machine learning system trained using the system of claim 1.

11. A system for training a machine learning system, comprising:

an expected value update component that, for a plurality of outputs and for a plurality of instances in which a single feature function is non-zero, modifies an expected value based, at least in part, upon the single feature function of an input vector and an output value, a sum of lambda variable and a normalization variable, modification of the expected value being based, at least in part, upon the following equation;

expected value=expected value+ƒ

_i({overscore (x)}_j, y)e^s[j,y]/z[j]

where ƒ

_i({overscore (x)}_j, y) is the feature function,

{overscore (x)}_j is the input vector,

y is the output value,

s[j,y] is the sum of lambda variable, and,

z[j] is the normalization variable;

an error calculator that calculates an error based, at least in part, upon the expected value and an observed value;

a parameter update component that modifies class trainable parameters or word trainable parameters based, at least in part, upon the error; and

,

a variable update component that, for the plurality of outputs and for the plurality of instances in which the feature function is non-zero, sequentially updates at least one of the sum of lambda variable and the normalization variable based, at least in part, upon the error.

12. The system of claim 11, the class trainable parameters being trained before the word trainable parameters are trained.

13. The system of claim 11, further comprising a training data store that stores at least one of the observed value and the input vector.

14. A method for training a machine learning system, comprising:

for each feature function, updating an expected value based, at least in part, upon a feature function of an input vector and an output value, a sum of lambda variable and a normalization variable;

for each feature function, calculating an error based, at least in part, upon the expected value and an observed value, the error calculation being based, at least in part, upon the following equation;

15. The method of claim 14, further comprising at least one of word clustering, smoothing and improved iterative scaling.

16. A method for training a machine learning system, comprising:

updating an expected value based, at least in part, upon a feature function of an input vector and an output value, a sum of lambda variable and a normalization variable, for each output, for each instance that the feature function is not zero;

calculating an error based, at least in part, upon the expected value and an observed value, the error calculation further employing, at least in part, the following equation;

17. The method of claim 16, further comprising at least one of the following acts:

performing general initialization;

resetting an expected value;

determining whether there are more outputs; and

,

determining whether there are more feature functions.

18. A computer implemented method for training a learning system, comprising the following computer executable acts:

training trainable class parameters based, at least in part, upon sequential conditional generalized iterative scaling an input vector, an output value, and calculating an error employing, at least in part, the following equation;

19. A computer readable medium storing computer executable components of a system facilitating training of a machine learning system, comprising:

an expected value update component that modifies an expected value for a plurality of outputs and for a plurality of instances in which a single feature function is non-zero based, at least in part, upon the single feature function of an input vector and an output value, a sum of lambda variable and a normalization variable;

an error calculator component that calculates an error based, at least in part, upon the expected value and an observed value;

a parameter update component that modifies a trainable parameter based, at least in part, upon the error; and

,

a variable update component that sequentially updates at least one of the sum of lambda variable and the normalization variable for the plurality of outputs and for the plurality of instances in which the feature function is non-zero based, at least in part, upon the error, the updating of the sum of lambda variable and the normalization variable being based, at least in part, upon the following equations;

z[j]=z[j]−

e^s[j,y]

s[j,y]=s[j,y]+δ

_i

z[j]=z[j]+e^s[j,y]

where s[j,y] is the sum of lambda variable,

z[j] is the normalization variable, and,

δ

_i is the error.

20. A training system for a machine learning system, comprising:

means for modifying an expected value for a plurality of outputs and for a plurality of instances in which a feature function is non-zero based, at least in part, upon the feature function of an input vector and an output value, a sum of lambda variable and a normalization variable;

means for calculating an error based, at least in part, upon the expected value and an observed value, the means for error calculation further employing, at least in part, the following equation;