Learning Fast and Slow: A Unified Batch/Stream Framework

Montiel, Jacob; Bifet, Albert; Losing, Viktor; Read, Jesse; Abdessalem, Talel

doi:10.1109/bigdata.2018.8622222

Cited by 21 publications

(25 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The proposed method is publicly available online to support reproducible science 4 . Our implementation is written in Python and is based on the scikit-multiflow framework [16].…”

Section: Contributions and Paper Organisationmentioning

confidence: 99%

Unsupervised Concept Drift Detection Using a Student–Teacher Approach

2020

View full text Add to dashboard Cite

Concept drift detection is a crucial task in data stream evolving environments. Most of the state of the art approaches designed to tackle this problem monitor the loss of predictive models. Accordingly, an alarm is launched when the loss increases significantly, which triggers some adaptation mechanism (e.g. retrain the model). However, this modus operandi falls short in many real-world scenarios, where the true labels are not readily available to compute the loss. These often take up to several weeks to be available. In this context, there is increasing attention to approaches that perform concept drift detection in an unsupervised manner, i.e., without access to the true labels. We propose a novel approach to unsupervised concept drift detection, which is based on a student-teacher learning paradigm. Essentially, we create an auxiliary model (student) to mimic the behaviour of the main model (teacher). At run-time, our approach is to use the teacher for predicting new instances and monitoring the mimicking loss of the student for concept drift detection. In a set of controlled experiments, we discovered that the proposed approach detects concept drift effectively. Relative to the gold standard, in which the labels are immediately available after prediction, our approach is more conservative: it signals less false alarms, but it requires more time to detect changes. We also show the competitiveness of our approach relative to other unsupervised methods.

show abstract

“…The proposed method is publicly available online to support reproducible science 4 . Our implementation is written in Python and is based on the scikit-multiflow framework [16].…”

Section: Contributions and Paper Organisationmentioning

confidence: 99%

Unsupervised Concept Drift Detection Using a Student–Teacher Approach

2020

View full text Add to dashboard Cite

show abstract

“…The Online Mondrian reaches state-of-the-art performances in the real datasets [12], [13], [14], [9]. The Data Stream Mondrian 2 GB achieves similar performances for the Banos et al dataset and RandomRBF stable.…”

Section: A Baselinesmentioning

confidence: 82%

Mondrian Forest for Data Stream Classification Under Memory Constraints

Khannouz¹,

Glatard²

2022

Preprint

View full text Add to dashboard Cite

Supervised learning algorithms generally assume the availability of enough memory to store their data model during the training and test phases. However, in the Internet of Things, this assumption is unrealistic when data comes in the form of infinite data streams, or when learning algorithms are deployed on devices with reduced amounts of memory. In this paper, we adapt the online Mondrian forest classification algorithm to work with memory constraints on data streams. In particular, we design five out-of-memory strategies to update Mondrian trees with new data points when the memory limit is reached. Moreover, we design trimming mechanisms to make Mondrian trees more robust to concept drifts under memory constraints. We evaluate our algorithms on a variety of real and simulated datasets, and we conclude with recommendations on their use in different situations: the Extend Node strategy appears as the best out-of-memory strategy in all configurations, whereas different trimming mechanisms should be adopted depending on whether a concept drift is expected. All our methods are implemented in the OrpailleCC open-source library and are ready to be used on embedded systems and connected objects.

show abstract

“…Over time the precision and recall converge to 88% and 85% respectively. These values have been achieved by using the default configuration of the scikit-multiflow library [20].…”

Section: Methodsmentioning

confidence: 99%

Evaluating Crowdshipping Systems with Agent-Based Simulation

Dötterl

Bruns

Dunkel

et al. 2020

Multi-Agent Systems and Agreement Technologies

View full text Add to dashboard Cite

In parcel delivery, the "last mile" from the parcel hub to the customer is costly, especially for time-sensitive delivery tasks that have to be completed within hours after arrival. Recently, crowdshipping has attracted increased attention as a new alternative to traditional delivery modes. In crowdshipping, private citizens ("the crowd") perform short detours in their daily lives to contribute to parcel delivery in exchange for small incentives. However, achieving desirable crowd behavior is challenging as the crowd is highly dynamic and consists of autonomous, self-interested individuals. Leveraging crowdshipping for time-sensitive deliveries remains an open challenge. In this paper, we present an agent-based approach to on-time parcel delivery with crowds. Our system performs data stream processing on the couriers' smartphone sensor data to predict delivery delays. Whenever a delay is predicted, the system attempts to forge an agreement for transferring the parcel from the current deliverer to a more promising courier nearby. Our experiments show that through accurate delay predictions and purposeful task transfers many delays can be prevented that would occur without our approach.

show abstract

Learning Fast and Slow: A Unified Batch/Stream Framework

Cited by 21 publications

References 18 publications

Unsupervised Concept Drift Detection Using a Student–Teacher Approach

Unsupervised Concept Drift Detection Using a Student–Teacher Approach

Mondrian Forest for Data Stream Classification Under Memory Constraints

Evaluating Crowdshipping Systems with Agent-Based Simulation

Contact Info

Product

Resources

About