While a range of computing equipment has been developed or proposed for use to solve machine learning problems in edge computing configurations, one of the least-explored options is the use of clusters of low-resource computing devices, such as the Raspberry Pi. Although such hardware configurations have been discussed in the past, their performance for ML tasks remains unexplored. In this paper, we discuss the performance of a Raspberry Pi micro-cluster, configured with industry-standard platforms, using Hadoop for distributed file storage and Spark for machine learning. Using the latest Raspberry Pi 4 model (quad core 1.5GHz, 4Gb RAM), we find encouraging results for use of such micro-clusters both for local training of ML models and execution of ML-based predictions. Our aim is to use such computing resources in a distributed architecture to serve tourism applications through the analysis of big data.

Komninos, A., Simou I., Gkorgkolis N., & Garofalakis J. (2019).  Performance of Raspberry Pi microclusters for Edge Machine Learning in Tourism. Edge Machine Learning for Smart IoT Environments Workshop (EDGING), 2019 European Conference on Ambient Intelligence (AmI2019). Rome, Italy, CEUR Workshop Proceedings.

We analyse data from a longitudinal study of 44participants, including notification handling, device state and location information. We demonstrate that it is possible to semantically label a user’s location based on their notification handling behaviour, even when location coordinates are obfuscated so as not to precisely match known venue lo- cations. Privacy-preserving semantic labelling of a user’s location can be useful for the contextually-relevant handling of interruptions and service delivery on mobile device

Komninos, A., Simou I., Frengkou E., & Garofalakis J. (2019).  Discovering User Location Semantics using Mobile Notification Handling Behaviour. 15th European Conference on Ambient Intelligence (AmI'19). Rome, Italy, Springer. DOI:10.1007/978-3-030-34255-5_15

This paper presents ongoing work in the development of a scalable hydroponics monitoring system. Our system leverages using wireless IoT technology and applies machine learning techniques on gath- ered data to provide recommendations to agronomists. Hydroponics is a method of growing plants in a water based nutrient rich solution system, instead of soil. By monitoring the parameters of the solution and the en- vironmental parameters inside the greenhouse, farmers can increase the production while decreasing the need for manual labor. Multiple net- worked sensors can measure these parameters and send all the necessary information to an Internet of things (IoT) platform (i.e., Thingsboard) in order the farmer to be able to control and adjust current operating conditions (e.g. environmental controls) and plan the nutrition schedule. Machine Learning can be used to detect anomalous operating conditions and to provide operational recommendations to assist farmers. The nov- elty presented in our system is that data contributed by multiple farming sites can be used to improve the quality of predictions and recommen- dations for all parties involved.

Georgiadis, G., Komninos A., Koskeris A., & Garofalakis J. (2019).  Improving Hydroponic Agriculture through IoT-enabled Collaborative Machine Learning. Intl. Workshop on Data Science and Internet of Things. Catania, Italy.

This paper presents the development of a scalable hydro- ponics monitoring system and data processing through wireless sensor networks. We implement novel hardware and virtual sensors through ab- straction in an IoT management platform. We introduce the concept of collaborative machine learning from multiple sites to improve prediction and discuss related project challenges

Komninos, A., Georgiadis G., Koskeris A., & Garofalakis J. (2019).  Internet of things applications on monitoring hydroponics through wireless sensor networks. 10th International Conference on Information, Intelligence, Systems and Applications (IISA'19). ., Springer.