The 33rd MLSP Workshop is organizing the Data Competition (DC) Sessions with new and exciting problems that aim at engaging machine learning and signal processing research communities, industries, and practitioners. DC sessions will be organized in the MLSP 2023 program. This session will include DC paper presentations (oral or poster), followed by a panel or open discussion.

The top 3 ranked teams will be invited to submit a (maximum) 6-page paper and present it at MLSP 2023. The accepted papers will be published in the MLSP proceedings.

Please note that the papers submitted to Data Competition will undergo the same review evaluation process of the regular papers submitted to MLSP 2023. Please email your inquiry related to special sessions to:

The following Data Competition have been accepted at IEEE MLSP 2023.

Urban Wireless Localization Competition

In urban environments, Global Navigation Satellite Systems fail to provide good accuracy because of the low likelihood of line-of-sight (LOS) links between the User Equipment (UE) to be located and the satellites, due to the presence the buildings that block the satellite signals. Thus, other approaches, that can reliably operate under non-line-of-sight (NLOS) conditions are required. In the recent years, many researchers have developed algorithms that are tailored to perform under heavy NLOS conditions. Many of such methods use Received Signal Strength (RSS) or Time of Arrival (ToA) (or both, in hybrid methods) information regarding the beacon signals that are regularly sent from wireless anchor infrastructure nodes such as Base Stations (BS) or Access Points (AP). In order to foster research and facilitate fair comparisons among the methods in realistic NLOS heavy conditions, we provide a pathloss radio map dataset based on accurate radio wave propagation simulations along with the corresponding novel ToA radio map dataset generated under realistic urban scenarios and launch the Urban Wireless Localization Competition.

The main task of the competition is to develop highly accurate localization methods which can make use of RSS (pathloss) and ToA measurements. The ranking of the methods is going to be based on the average of their accuracies in NLOS-only and NLOS/LOS-mixed scenarios and under different noise levels on the RSS and ToA measurements.

Organizers: Çağkan Yapar, Fabian Jaensch, Ron Levie, Gitta Kutyniok, Giuseppe Caire (Technische Universität Berlin, Germany)

Data Competition Website:


Volunteer Retention and Future Collaboration Prediction Challenge

In recent years, online volunteer crowdsourcing platforms have emerged, facilitating collaborative efforts to address community needs, particularly during crises such as earthquakes and pandemics. To optimize these efforts, it’s crucial to model volunteers’ task participation and collaboration behaviors. In this challenge, we aim to learn such models using task participation data from a mobile volunteer organization platform. The dataset consists of volunteer activity records from 20,000 users during 2020-2022.

Our specific goals are to:

1) Predict volunteers’ retention based on their participation history,

2) Predict future collaborations between pairs of volunteers.

A training dataset is available upon request.

Organizers: Shutong Chen, Yang Li (Tsinghua-Berkeley Shenzhen Institute, China)

Data Competition Website: