Method and apparatus for machine learning

• US 6,249,781 B1
• Filed: 05/05/1999
• Issued: 06/19/2001
• Est. Priority Date: 09/18/1992
• Status: Expired due to Fees

• Current Assignee: Verizon Laboratories Incorporated (Verizon Communications Inc.)
• Original Assignee: Verizon Laboratories Incorporated (Verizon Communications Inc.)
• Inventors: Sutton, Richard S.
• Primary Examiner(s): Chaki, Kakali
• Assistant Examiner(s): STARKS, WILBERT L
• Application Number:
  US09/305,575
• Time in Patent Office: 776 Days
• Field of Search: 706/25, 706/21, 706/23, 706/27
• US Class Current: 706/25
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

A method and apparatus is disclosed for machine learning of a pattern sequence which is derived from a plurality of inputs. The pattern sequence is predicted from learning rate parameters that are exponentially related to an incrementally calculated gain parameter for each input. The gain parameter are increased or decreased in real time in correlation with the accuracy of the learning process. The disclosed method and apparatus are advantageously utilized in signal processing, adaptive control systems, and pattern recognition.

Claims

1. A computer system for determining a time dependent pattern sequence y(t), comprising:

a memory configured to store instructions; and

a processor configured to execute the instructions to receive a plurality of time dependent inputs x_i(t) and a meta-step-size parameter θ

, determine a predicted value of the pattern sequence y(t) from the time dependent inputs based on a learning rate α

_i that is exponentially related to an incremental gain β

_i(t), the incremental gain β

_i(t) being derived from previous values of β

_i(t),

store the predicted value of the pattern sequence y(t) in the memory, and determine the pattern sequence y(t) using the stored predicted value.

2. The computer system of claim 1, wherein the predicted value of the pattern sequence is calculated as a linear combination of the plurality of time dependent inputs.

3. The computer system of claim 1, wherein the derivation means derives the incremental gain β

_i(t) according to the rule;

4. The computer system of claim 1, wherein the predicted value of the pattern sequence is calculated as a non-linear combination of the plurality of time dependent inputs.

5. The computer system of claim 1, wherein the derivation means derives the incremental gain β

_i(t) according to the rule;