Towards population scale activity recognition: A framework for handling data diversity

Saeed Abdullah, Nicholas D. Lane, Tanzeem Choudhury

Research output: Chapter in Book/Report/Conference proceedingConference contribution

13 Citations (Scopus)

Abstract

The rising popularity of the sensor-equipped smartphone is changing the possible scale and scope of human activity inference. The diversity in user population seen in large user bases can overwhelm conventional one-size-fits-all classication approaches. Although personalized models are better able to handle population diversity, they often require increased effort from the end user during training and are computationally expensive. In this paper, we propose an activity classification framework that is scalable and can tractably handle an increasing number of users. Scalability is achieved by maintaining distinct groups of similar users during the training process, which makes it possible to account for the differences between users without resorting to training individualized classifiers. The proposed framework keeps user burden low by leveraging crowd-sourced data labels, where simple natural language processing techniques in combination with multiinstance learning are used to handle labeling errors introduced by low-commitment everyday users. Experiment results on a large public dataset demonstrate that the framework can cope with population diversity irrespective of population size.

Original languageEnglish (US)
Title of host publicationAAAI-12 / IAAI-12 - Proceedings of the 26th AAAI Conference on Artificial Intelligence and the 24th Innovative Applications of Artificial Intelligence Conference
Pages851-857
Number of pages7
StatePublished - Nov 7 2012
Event26th AAAI Conference on Artificial Intelligence and the 24th Innovative Applications of Artificial Intelligence Conference, AAAI-12 / IAAI-12 - Toronto, ON, Canada
Duration: Jul 22 2012Jul 26 2012

Publication series

NameProceedings of the National Conference on Artificial Intelligence
Volume2

Other

Other26th AAAI Conference on Artificial Intelligence and the 24th Innovative Applications of Artificial Intelligence Conference, AAAI-12 / IAAI-12
CountryCanada
CityToronto, ON
Period7/22/127/26/12

Fingerprint

Data handling
Smartphones
Labeling
Scalability
Labels
Classifiers
Sensors
Processing
Experiments

All Science Journal Classification (ASJC) codes

  • Software
  • Artificial Intelligence

Cite this

Abdullah, S., Lane, N. D., & Choudhury, T. (2012). Towards population scale activity recognition: A framework for handling data diversity. In AAAI-12 / IAAI-12 - Proceedings of the 26th AAAI Conference on Artificial Intelligence and the 24th Innovative Applications of Artificial Intelligence Conference (pp. 851-857). (Proceedings of the National Conference on Artificial Intelligence; Vol. 2).
Abdullah, Saeed ; Lane, Nicholas D. ; Choudhury, Tanzeem. / Towards population scale activity recognition : A framework for handling data diversity. AAAI-12 / IAAI-12 - Proceedings of the 26th AAAI Conference on Artificial Intelligence and the 24th Innovative Applications of Artificial Intelligence Conference. 2012. pp. 851-857 (Proceedings of the National Conference on Artificial Intelligence).
@inproceedings{5f4e73eab01b476ab4a8c455e3bd30e0,
title = "Towards population scale activity recognition: A framework for handling data diversity",
abstract = "The rising popularity of the sensor-equipped smartphone is changing the possible scale and scope of human activity inference. The diversity in user population seen in large user bases can overwhelm conventional one-size-fits-all classication approaches. Although personalized models are better able to handle population diversity, they often require increased effort from the end user during training and are computationally expensive. In this paper, we propose an activity classification framework that is scalable and can tractably handle an increasing number of users. Scalability is achieved by maintaining distinct groups of similar users during the training process, which makes it possible to account for the differences between users without resorting to training individualized classifiers. The proposed framework keeps user burden low by leveraging crowd-sourced data labels, where simple natural language processing techniques in combination with multiinstance learning are used to handle labeling errors introduced by low-commitment everyday users. Experiment results on a large public dataset demonstrate that the framework can cope with population diversity irrespective of population size.",
author = "Saeed Abdullah and Lane, {Nicholas D.} and Tanzeem Choudhury",
year = "2012",
month = "11",
day = "7",
language = "English (US)",
isbn = "9781577355687",
series = "Proceedings of the National Conference on Artificial Intelligence",
pages = "851--857",
booktitle = "AAAI-12 / IAAI-12 - Proceedings of the 26th AAAI Conference on Artificial Intelligence and the 24th Innovative Applications of Artificial Intelligence Conference",

}

Abdullah, S, Lane, ND & Choudhury, T 2012, Towards population scale activity recognition: A framework for handling data diversity. in AAAI-12 / IAAI-12 - Proceedings of the 26th AAAI Conference on Artificial Intelligence and the 24th Innovative Applications of Artificial Intelligence Conference. Proceedings of the National Conference on Artificial Intelligence, vol. 2, pp. 851-857, 26th AAAI Conference on Artificial Intelligence and the 24th Innovative Applications of Artificial Intelligence Conference, AAAI-12 / IAAI-12, Toronto, ON, Canada, 7/22/12.

Towards population scale activity recognition : A framework for handling data diversity. / Abdullah, Saeed; Lane, Nicholas D.; Choudhury, Tanzeem.

AAAI-12 / IAAI-12 - Proceedings of the 26th AAAI Conference on Artificial Intelligence and the 24th Innovative Applications of Artificial Intelligence Conference. 2012. p. 851-857 (Proceedings of the National Conference on Artificial Intelligence; Vol. 2).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Towards population scale activity recognition

T2 - A framework for handling data diversity

AU - Abdullah, Saeed

AU - Lane, Nicholas D.

AU - Choudhury, Tanzeem

PY - 2012/11/7

Y1 - 2012/11/7

N2 - The rising popularity of the sensor-equipped smartphone is changing the possible scale and scope of human activity inference. The diversity in user population seen in large user bases can overwhelm conventional one-size-fits-all classication approaches. Although personalized models are better able to handle population diversity, they often require increased effort from the end user during training and are computationally expensive. In this paper, we propose an activity classification framework that is scalable and can tractably handle an increasing number of users. Scalability is achieved by maintaining distinct groups of similar users during the training process, which makes it possible to account for the differences between users without resorting to training individualized classifiers. The proposed framework keeps user burden low by leveraging crowd-sourced data labels, where simple natural language processing techniques in combination with multiinstance learning are used to handle labeling errors introduced by low-commitment everyday users. Experiment results on a large public dataset demonstrate that the framework can cope with population diversity irrespective of population size.

AB - The rising popularity of the sensor-equipped smartphone is changing the possible scale and scope of human activity inference. The diversity in user population seen in large user bases can overwhelm conventional one-size-fits-all classication approaches. Although personalized models are better able to handle population diversity, they often require increased effort from the end user during training and are computationally expensive. In this paper, we propose an activity classification framework that is scalable and can tractably handle an increasing number of users. Scalability is achieved by maintaining distinct groups of similar users during the training process, which makes it possible to account for the differences between users without resorting to training individualized classifiers. The proposed framework keeps user burden low by leveraging crowd-sourced data labels, where simple natural language processing techniques in combination with multiinstance learning are used to handle labeling errors introduced by low-commitment everyday users. Experiment results on a large public dataset demonstrate that the framework can cope with population diversity irrespective of population size.

UR - http://www.scopus.com/inward/record.url?scp=84868279929&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84868279929&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84868279929

SN - 9781577355687

T3 - Proceedings of the National Conference on Artificial Intelligence

SP - 851

EP - 857

BT - AAAI-12 / IAAI-12 - Proceedings of the 26th AAAI Conference on Artificial Intelligence and the 24th Innovative Applications of Artificial Intelligence Conference

ER -

Abdullah S, Lane ND, Choudhury T. Towards population scale activity recognition: A framework for handling data diversity. In AAAI-12 / IAAI-12 - Proceedings of the 26th AAAI Conference on Artificial Intelligence and the 24th Innovative Applications of Artificial Intelligence Conference. 2012. p. 851-857. (Proceedings of the National Conference on Artificial Intelligence).