Machine Teaching of Active Sequential Learners

Tutkimustuotos: Artikkeli kirjassa/konferenssijulkaisussavertaisarvioitu

Standard

Machine Teaching of Active Sequential Learners. / Peltola, Tomi; Çelikok, Mustafa Mert; Daee, Pedram; Kaski, Samuel.

33rd Conference on Neural Information Processing Systems: NeurIPS 2019 . Neural Information Processing Systems Foundation, 2019. s. 11202-11213 (Advances in Neural Information Processing Systems).

Tutkimustuotos: Artikkeli kirjassa/konferenssijulkaisussavertaisarvioitu

Harvard

Peltola, T, Çelikok, MM, Daee, P & Kaski, S 2019, Machine Teaching of Active Sequential Learners. julkaisussa 33rd Conference on Neural Information Processing Systems: NeurIPS 2019 . Advances in Neural Information Processing Systems, Neural Information Processing Systems Foundation, Sivut 11202-11213, Vancouver, Kanada, 08/12/2019.

APA

Peltola, T., Çelikok, M. M., Daee, P., & Kaski, S. (2019). Machine Teaching of Active Sequential Learners. teoksessa 33rd Conference on Neural Information Processing Systems: NeurIPS 2019 (Sivut 11202-11213). (Advances in Neural Information Processing Systems). Neural Information Processing Systems Foundation.

Vancouver

Peltola T, Çelikok MM, Daee P, Kaski S. Machine Teaching of Active Sequential Learners. julkaisussa 33rd Conference on Neural Information Processing Systems: NeurIPS 2019 . Neural Information Processing Systems Foundation. 2019. s. 11202-11213. (Advances in Neural Information Processing Systems).

Author

Peltola, Tomi ; Çelikok, Mustafa Mert ; Daee, Pedram ; Kaski, Samuel. / Machine Teaching of Active Sequential Learners. 33rd Conference on Neural Information Processing Systems: NeurIPS 2019 . Neural Information Processing Systems Foundation, 2019. Sivut 11202-11213 (Advances in Neural Information Processing Systems).

Bibtex - Lataa

@inproceedings{d3b5c57074104e4f8b204e9ad4e1582d,
title = "Machine Teaching of Active Sequential Learners",
abstract = "Machine teaching addresses the problem of finding the best training data that can guide a learning algorithm to a target model with minimal effort. In conventional settings, a teacher provides data that are consistent with the true data distribution. However, for sequential learners which actively choose their queries, such as multi-armed bandits and active learners, the teacher can only provide responses to the learner’s queries, not design the full data. In this setting, consistent teachers can be sub-optimal for finite horizons. We formulate this sequential teaching problem, which current techniques in machine teaching do not address, as a Markov decision process, with the dynamics nesting a model of the learner and the actions being the teacher's responses. Furthermore, we address the complementary problem of learning from a teacher that plans: to recognise the teaching intent of the responses, the learner is endowed with a model of the teacher. We test the formulation with multi-armed bandit learners in simulated experiments and a user study. The results show that learning is improved by (i) planning teaching and (ii) the learner having a model of the teacher. The approach gives tools to taking into account strategic (planning) behaviour of users of interactive intelligent systems, such as recommendation engines, by considering them as boundedly optimal teachers.",
author = "Tomi Peltola and {\cC}elikok, {Mustafa Mert} and Pedram Daee and Samuel Kaski",
note = "Kokous on 8.-14.12.2019. Avaa tiedosto sitten joskus, kun proceedings ilmestyy. Etsi ISBN tai muuta D-luokkaan.",
year = "2019",
month = "11",
day = "28",
language = "English",
series = "Advances in Neural Information Processing Systems",
publisher = "Neural Information Processing Systems Foundation",
pages = "11202--11213",
booktitle = "33rd Conference on Neural Information Processing Systems",

}

RIS - Lataa

TY - GEN

T1 - Machine Teaching of Active Sequential Learners

AU - Peltola, Tomi

AU - Çelikok, Mustafa Mert

AU - Daee, Pedram

AU - Kaski, Samuel

N1 - Kokous on 8.-14.12.2019. Avaa tiedosto sitten joskus, kun proceedings ilmestyy. Etsi ISBN tai muuta D-luokkaan.

PY - 2019/11/28

Y1 - 2019/11/28

N2 - Machine teaching addresses the problem of finding the best training data that can guide a learning algorithm to a target model with minimal effort. In conventional settings, a teacher provides data that are consistent with the true data distribution. However, for sequential learners which actively choose their queries, such as multi-armed bandits and active learners, the teacher can only provide responses to the learner’s queries, not design the full data. In this setting, consistent teachers can be sub-optimal for finite horizons. We formulate this sequential teaching problem, which current techniques in machine teaching do not address, as a Markov decision process, with the dynamics nesting a model of the learner and the actions being the teacher's responses. Furthermore, we address the complementary problem of learning from a teacher that plans: to recognise the teaching intent of the responses, the learner is endowed with a model of the teacher. We test the formulation with multi-armed bandit learners in simulated experiments and a user study. The results show that learning is improved by (i) planning teaching and (ii) the learner having a model of the teacher. The approach gives tools to taking into account strategic (planning) behaviour of users of interactive intelligent systems, such as recommendation engines, by considering them as boundedly optimal teachers.

AB - Machine teaching addresses the problem of finding the best training data that can guide a learning algorithm to a target model with minimal effort. In conventional settings, a teacher provides data that are consistent with the true data distribution. However, for sequential learners which actively choose their queries, such as multi-armed bandits and active learners, the teacher can only provide responses to the learner’s queries, not design the full data. In this setting, consistent teachers can be sub-optimal for finite horizons. We formulate this sequential teaching problem, which current techniques in machine teaching do not address, as a Markov decision process, with the dynamics nesting a model of the learner and the actions being the teacher's responses. Furthermore, we address the complementary problem of learning from a teacher that plans: to recognise the teaching intent of the responses, the learner is endowed with a model of the teacher. We test the formulation with multi-armed bandit learners in simulated experiments and a user study. The results show that learning is improved by (i) planning teaching and (ii) the learner having a model of the teacher. The approach gives tools to taking into account strategic (planning) behaviour of users of interactive intelligent systems, such as recommendation engines, by considering them as boundedly optimal teachers.

UR - https://aaltopml.github.io/machine-teaching-of-active-sequential-learners/

M3 - Conference contribution

T3 - Advances in Neural Information Processing Systems

SP - 11202

EP - 11213

BT - 33rd Conference on Neural Information Processing Systems

PB - Neural Information Processing Systems Foundation

ER -

ID: 36741353