17551746. MEMOIZING MACHINE-LEARNING PRE-PROCESSING AND FEATURE ENGINEERING simplified abstract (INTERNATIONAL BUSINESS MACHINES CORPORATION)

From WikiPatents
Jump to navigation Jump to search

MEMOIZING MACHINE-LEARNING PRE-PROCESSING AND FEATURE ENGINEERING

Organization Name

INTERNATIONAL BUSINESS MACHINES CORPORATION

Inventor(s)

TAKUYA Nakaike of Yokohama-shi (JP)

Motohiro Kawahito of Sagamihara-shi (JP)

MEMOIZING MACHINE-LEARNING PRE-PROCESSING AND FEATURE ENGINEERING - A simplified explanation of the abstract

This abstract first appeared for US patent application 17551746 titled 'MEMOIZING MACHINE-LEARNING PRE-PROCESSING AND FEATURE ENGINEERING

Simplified Explanation

The abstract describes a method for creating a table of keys and values in a machine-learning pre-processing pipeline. The method measures various metrics such as hit rate, lookup time, execution time, and threshold on the number of input elements.

  • The method creates a table with keys and values based on an input array in a machine-learning pre-processing pipeline.
  • It measures the hit rate to the table, average lookup time, average execution time, and a threshold on the number of input elements.
  • When the conditions are met, the method looks up the value in the table using an element of the input array as a key.
  • If the value is not found in the table, the method calls the pipeline for the remaining elements in the input array.

Potential Applications

  • This method can be applied in various machine-learning pre-processing pipelines where efficient lookup and execution times are crucial.
  • It can be used in systems that require quick access to pre-processed data based on input elements.

Problems Solved

  • The method solves the problem of efficiently looking up values in a table within a machine-learning pre-processing pipeline.
  • It addresses the issue of reducing execution time by utilizing pre-processed values from the table.

Benefits

  • The method improves the efficiency of the machine-learning pre-processing pipeline by reducing lookup and execution times.
  • It allows for faster access to pre-processed data, improving overall system performance.


Original Abstract Submitted

A method creates a table of keys and values. Each key is an element of an input array which is an input of a machine-learning pre-processing pipeline, and each value is an output of the pipeline. The method measures (1) a hit rate H to the memo table, (2) an average time T to look up the table, (3) an average time T to execute the pipeline, and (4) a threshold T on a number of elements of the input array. The method looks up the value in the table by using an element of the input array as a key when T × H > T and the number of elements in the input array is less than T. The method calls the pipeline in place of the lookup for all of the remaining elements in the input array when the value is not in the table.

Retrieved from "http://wikipatents.org/index.php?title=17551746._MEMOIZING_MACHINE-LEARNING_PRE-PROCESSING_AND_FEATURE_ENGINEERING_simplified_abstract_(INTERNATIONAL_BUSINESS_MACHINES_CORPORATION)&oldid=26221"