MLUSD 2014 - 1st International Workshop on Machine Learning for Urban Sensor Data

As the focus in the Wireless Sensor Networks and Sensor Systems community is shifting from “How do we collect data?” to “What can we learn from the data and how do the models look like?” we want to bring researchers from this community and the Machine Learning community together. Working with sensor data, machine learning methods become more and more popular (e.g., at the ACM SenSys conference ? the major conference in this area ? in 2013 the First International Workshop on Sensing and Big Data Mining (SenseMine) took place).
As the applications for machine learning expand into other areas, the need for high-quality machine learning methods constantly grows. Additionally, there is a need for interpretable models as researchers want to grasp the models and get a sense of how the sensor information is combined in the model.
However, sensor data poses a number of unique challenges for machine learning. Ranging from missing values, unreliable measurements, missing calibration to high spatial diversity. Most challenges have not been addressed with a focus on real-world sensor data. It is our belief that a discussion will help foster new results in the intersection of both communities.
Topic Areas of Interest
Real-time machine learning
Iterative machine learning
Multi-target learning
Generating data analysis pipelines
Evaluation of machine learning models tailored to sensor data
Data extraction from sensor networks
Data conversion and calibration issues
Meta-learning, e.g., learning to adjust the analysis pipeline automatically
Interpretable models, e.g., Rule Learning or Decision Tree Learning
Generating high-quality data sets
Data quality issues
Dealing with missing and low quality data
Feature Engineering with a focus on sensor data features
Feature weighting and combination
Generating high-quality features from sensor data