SLRProp: A Back-Propagation Variant of Sparse Low Rank Method for DNNs Reduction.

Application of deep neural networks (DNN) in edge computing has emerged as a consequence of the need of real time and distributed response of different devices in a large number of scenarios. To this end, shredding these original structures is urgent due to the high number of parameters needed to represent them. As a consequence, the most representative components of different layers are kept in order to maintain the network's accuracy as close as possible to the entire network's ones. To do so, two different approaches have been developed in this work. First, the Sparse Low Rank Method (SLR) has been applied to two different Fully Connected (FC) layers to watch their effect on the final response, and the method has been applied to the latest of these layers as a duplicate. On the contrary, SLRProp has been proposed as a variant case, where the relevances of the previous FC layer's components were weighed as the sum of the products of each of these neurons' absolute values and the relevances of the neurons from the last FC layer that are connected with the neurons from the previous FC layer. Thus, the relationship of relevances across layer was considered. Experiments have been carried out in well-known architectures to conclude whether the relevances throughout layers have less effect on the final response of the network than the independent relevances intra-layer.

A Generalization Performance Study Using Deep Learning Networks in Embedded Systems.

Deep learning techniques are being increasingly used in the scientific community as a consequence of the high computational capacity of current systems and the increase in the amount of data available as a result of the digitalisation of society in general and the industrial world in particular. In addition, the immersion of the field of edge computing, which focuses on integrating artificial intelligence as close as possible to the client, makes it possible to implement systems that act in real time without the need to transfer all of the data to centralised servers. The combination of these two concepts can lead to systems with the capacity to make correct decisions and act based on them immediately and in situ. Despite this, the low capacity of embedded systems greatly hinders this integration, so the possibility of being able to integrate them into a wide range of micro-controllers can be a great advantage. This paper contributes with the generation of an environment based on Mbed OS and TensorFlow Lite to be embedded in any general purpose embedded system, allowing the introduction of deep learning architectures. The experiments herein prove that the proposed system is competitive if compared to other commercial systems.

Non-Invasive Ambient Intelligence in Real Life: Dealing with Noisy Patterns to Help Older People.

This paper aims to contribute to the field of ambient intelligence from the perspective of real environments, where noise levels in datasets are significant, by showing how machine learning techniques can contribute to the knowledge creation, by promoting software sensors. The created knowledge can be actionable to develop features helping to deal with problems related to minimally labelled datasets. A case study is presented and analysed, looking to infer high-level rules, which can help to anticipate abnormal activities, and potential benefits of the integration of these technologies are discussed in this context. The contribution also aims to analyse the usage of the models for the transfer of knowledge when different sensors with different settings contribute to the noise levels. Finally, based on the authors' experience, a framework proposal for creating valuable and aggregated knowledge is depicted.

Smart Meeting Room Usage Information and Prediction by Modelling Occupancy Profiles.

The monitoring of small houses and rooms has become possible due to the advances in IoT sensors, actuators and low power communication protocols in the last few years. As buildings are one of the biggest energy consuming entities, monitoring them has great interest for trying to avoid non-necessary energy waste. Moreover, human behaviour has been reported as being the main discrepancy source between energy usage simulations and real usage, so the ability to monitor and predict actions as opening windows, using rooms, etc. is gaining attention to develop stronger models which may lead to reduce the overall energy consumption of buildings, considering buildings thermal inertia and additional capabilities. In this paper, a case study is described in which four meeting rooms have been monitored to obtain information about the usage of the rooms and later use it to predict their future usage. The results show the possibility to deploy a simple and non-intrusive sensing system whose output could be used to develop advanced control strategies.

Development of a novel splice array platform and its application in the identification of alternative splice variants in lung cancer.

BACKGROUND: Microarrays strategies, which allow for the characterization of thousands of alternative splice forms in a single test, can be applied to identify differential alternative splicing events. In this study, a novel splice array approach was developed, including the design of a high-density oligonucleotide array, a labeling procedure, and an algorithm to identify splice events. RESULTS: The array consisted of exon probes and thermodynamically balanced junction probes. Suboptimal probes were tagged and considered in the final analysis. An unbiased labeling protocol was developed using random primers. The algorithm used to distinguish changes in expression from changes in splicing was calibrated using internal non-spliced control sequences. The performance of this splice array was validated with artificial constructs for CDC6, VEGF, and PCBP4 isoforms. The platform was then applied to the analysis of differential splice forms in lung cancer samples compared to matched normal lung tissue. Overexpression of splice isoforms was identified for genes encoding CEACAM1, FHL-1, MLPH, and SUSD2. None of these splicing isoforms had been previously associated with lung cancer. CONCLUSIONS: This methodology enables the detection of alternative splicing events in complex biological samples, providing a powerful tool to identify novel diagnostic and prognostic biomarkers for cancer and other pathologies.