In-Network Convolution in Grid Shaped Sensor Networks

Gathering information is the primary purpose of a Sensor Network. The task is performed by spatially distributed nodes equipped with sensing, processing, and communication capabilities. However, data gathered from a sensor network must be processed, and often the collective computation capability of nodes forming the sensor network is neglected in favor of data processing on cloud systems. Nowadays, Edge Computing has emerged as a new paradigm aiming to migrate data processing close to data sources. In this contribution, we focus on the development of a sensor network designed to detect a person’s fall. We named this sensor network the smart floor. Fall detection is tackled with a Convolutional Neural Network, and we propose an approach for in-network processing of convolution layers on grid-shaped sensor networks. The proposed approach could lead to the development of a sensor network that detects falls by performing CNN inference processing on the edge. We complement our work with a simulation using the simulator ns-3. The simulation is designed to emulate the communication overhead of the proposed approach applied to a wired sensor network that resembles the smart floor. Simulation results provide evidence on the feasibility of the proposed concept applied to wired grid shaped sensor networks.

Addressing Extreme Imbalance for Detecting Medications Mentioned in Twitter User Timelines

Tweets mentioning medications are valuable for efforts in digital epidemiology to supplement traditional methods of monitoring public health. A major obstacle, however, is to differentiate them from the large majority of tweets on other topics posted in a user’s timeline: solving the infamous ‘needle in a haystack’ problem. While deep learning models have significantly improved classification, their performance and inference processing time remain low on extremely imbalanced corpora where the tweets of interest are less than 1% of all tweets. In this study, we empirically evaluate under-sampling, fine-tuning, and filtering heuristics to train such classifiers. Using a corpus of 212 Twitter timelines (181,607 tweets with only 0.2% tweets mentioning a medication), our results show that combining these heuristics is necessary to impact the classifier’s performance. In our intrinsic evaluation, a classifier based on a lexicon and a BERT-base neural network achieved a 0.838 F1-score, a score similar to the ones of the best existing classifier, but it processed the corpus 28 times faster - a positive result, since processing speed is still a roadblock to deploying classifiers on large cohorts of Twitter users needed for pharmacovigilance. In our extrinsic evaluation, our classifier helped a labeler to extract the spans of medications more accurately and achieved a 0.76 Strict F1-score. To the best of our knowledge, this is the first evaluation of medications extraction in Twitter timelines and it establishes the first benchmark for future studies.