Robust evaluation of deep learning-based representation methods for survival and gene essentiality prediction on bulk RNA-seq data.

Deep learning (DL) has shown potential to provide powerful representations of bulk RNA-seq data in cancer research. However, there is no consensus regarding the impact of design choices of DL approaches on the performance of the learned representation, including the model architecture, the training methodology and the various hyperparameters. To address this problem, we evaluate the performance of various design choices of DL representation learning methods using TCGA and DepMap pan-cancer datasets and assess their predictive power for survival and gene essentiality predictions. We demonstrate that baseline methods achieve comparable or superior performance compared to more complex models on survival predictions tasks. DL representation methods, however, are the most efficient to predict the gene essentiality of cell lines. We show that auto-encoders (AE) are consistently improved by techniques such as masking and multi-head training. Our results suggest that the impact of DL representations and of pretraining are highly task- and architecture-dependent, highlighting the need for adopting rigorous evaluation guidelines. These guidelines for robust evaluation are implemented in a pipeline made available to the research community.

Multi-modal Biometrics Based Implicit Driver Identification System Using Multi-TF Images of ECG and EMG.

As security is emphasized inside and outside the vehicle, research on driver identification technology using bio-signals is being actively studied. The bio-signals acquired by the behavioral characteristics of the driver include artifacts generated according to the driving environment, which could potentially degrade the accuracy of the identification system. Existing driver identification systems either remove the normalization process of bio-signals in the preprocessing stage or use artifacts included in a single bio-signals, resulting in low identification accuracy. To solve these problems in a real situation, we propose a driver identification system that converts ECG and EMG signals obtained from different driving conditions into 2D spectrograms through multi-TF image and uses multi-stream CNN. The proposed system consists of a preprocessing phase of ECG and EMG signals, a multi-TF image conversion process, and a driver identification stage using a multi-stream-based CNN. Under all driving conditions, the driver identification system reached an average accuracy of 96.8% and an F1 score of 0.973, which overperformed the existing driver identification systems by more than 1%.

Interoperable Blockchains for Highly-Integrated Supply Chains in Collaborative Manufacturing.

Blockchain technology plays a pivotal role in the undergoing fourth industrial revolution or Industry 4.0. It is considered a tremendous boost to company digitalization; thus, considerable investments in blockchain are being made. However, there is no single blockchain technology, but various solutions exist, and they cannot interoperate with one each other. The ecosystem envisioned by the Industry 4.0 does not have centralized management or leading organization, so a single blockchain solution cannot be imposed. The various organizations hold their own blockchains, which must interoperate seamlessly. Despite some solutions for blockchain interoperability being proposed, the problem is still open. This paper aims to devise a secure solution for blockchain interoperability. The proposed approach consists of a relay scheme based on Trusted Execution Environment to provide higher security guarantees than the current literature. In particular, the proposed solution adopts an off-chain secure computation element invoked by a smart contract on a blockchain to securely communicate with its peered counterpart. A prototype has been implemented and used for the performance assessment, e.g., to measure the latency increase due to cross-blockchain interactions. The achieved and reported experimental results show that the proposed security solution introduces an additional latency that is entirely tolerable for transactions. At the same time, the usage of the Trusted Execution Environment imposes a negligible overhead.

Semi-automated transcription and scoring of autobiographical memory narratives.

Autobiographical memory studies conducted with narrative methods are onerous, requiring significant resources in time and labor. We have created a semi-automated process that allows autobiographical transcribing and scoring methods to be streamlined. Our paper focuses on the Autobiographical Interview (AI; Levine, Svoboda, Hay, Winocur, & Moscovitch, Psychology and Aging, 17, 677-89, 2002), but this method can be adapted for other narrative protocols. Specifically, here we lay out a procedure that guides researchers through the four main phases of the autobiographical narrative pipeline: (1) data collection, (2) transcribing, (3) scoring, and (4) analysis. First, we provide recommendations for incorporating transcription software to augment human transcribing. We then introduce an electronic scoring procedure for tagging narratives for scoring that incorporates the traditional AI scoring method with basic keyboard shortcuts in Microsoft Word. Finally, we provide a Python script that can be used to automate counting of scored transcripts. This method accelerates the time it takes to conduct a narrative study and reduces the opportunity for error in narrative quantification. Available open access on GitHub ( https://github.com/cMadan/scoreAI ), our pipeline makes narrative methods more accessible for future research.

A Night at the OPERA: A Conceptual Framework for an Integrated Distributed Sensor Network-Based System to Figure out Safety Protocols for Animals under Risk of Fire.

Large scale wildfire events that occurred around the world involved a massive loss of animal lives, with a consequent economic impact on agricultural holdings and damages to ecosystems. Preparing animals for a wildfire evacuation requires an extra level of planning, preparedness and coordination, which is missing in the current practice. This paper describes a conceptual framework of an ICT system implemented to support the activities of the Regional Veterinary referral Center for non-epidemic emergencies (CeRVEnE) in the Campania Region for the twofold objectives. On the one hand, it realizes the monitoring of the wooded areas under risk of fire in the so-called "Mount Vesuvius' red zone". On the other hand, it determines the OPtimal Evacuation Route for Animals (OPERA) in case of fire, for each of the reported animal species living in the mentioned red zone. The main innovation of the proposed system lies in its software architecture that aims at integrating a Distributed Sensor Network (DSN), an ad-hoc software to generate timely simulations for fire risk modeling, and a GIS (Geographic Information System) for both the activities of web mapping and OPERA definition. This paper shows some effective preliminary results of the system implementation. The importance of the system mainly lies in its accordance with the so-called "Foresight approach" perspective, that provides models and tools to guarantee the prevention of systematic failure in disaster risk management, and becomes moreover critical in the case of Mount Vesuvius, which hosts a unique combination of both animal and anthropic elements within a delicate natural ecosystem.