Q-Learning Sine-Cosine Algorithm (QLSCA) 
The sine-cosine algorithm (SCA) is a brand new population-based meta-heuristic algorithm. Along with exploiting sine and cosine features to carry out native and international searches (therefore the title sine-cosine), the SCA introduces a number of random and adaptive parameters to facilitate the search course of. Though it exhibits promising outcomes, the search technique of the SCA is weak to native minima/maxima because of the adoption of a hard and fast swap likelihood and the bounded magnitude of the sine and cosine features (from -1 to 1). On this paper, we suggest a brand new hybrid Q-learning sine-cosine- primarily based technique, known as the Q-learning sine-cosine algorithm (QLSCA). Inside the QLSCA, we eradicate the switching likelihood. As an alternative, we depend on the Q-learning algorithm (primarily based on the penalty and reward mechanism) to dynamically determine one of the best operation throughout runtime. Moreover, we combine two new operations (L’evy flight movement and crossover) into the QLSCA to facilitate leaping out of native minima/maxima and improve the answer range. To evaluate its efficiency, we undertake the QLSCA for the combinatorial check suite minimization downside. Experimental outcomes reveal that the QLSCA is statistically superior with regard to check suite measurement discount in comparison with latest state-of-the-art methods, together with the unique SCA, the particle swarm check generator (PSTG), adaptive particle swarm optimization (APSO) and the cuckoo search technique (CS) on the 95% confidence stage. Nevertheless, regarding the comparability with discrete particle swarm optimization (DPSO), there isn’t any vital distinction in efficiency on the 95% confidence stage. On a constructive be aware, the QLSCA statistically outperforms the DPSO in sure configurations on the 90% confidence stage. …
Cyber-Physical System (CPS)
A cyber-physical (additionally styled cyberphysical) system (CPS) is a mechanism that’s managed or monitored by computer-based algorithms, tightly built-in with the Web and its customers. In cyber-physical programs, bodily and software program elements are deeply intertwined, every working on totally different spatial and temporal scales, exhibiting a number of and distinct behavioral modalities, and interacting with one another in loads of ways in which change with context. Examples of CPS embrace sensible grid, autonomous car programs, medical monitoring, course of management programs, robotics programs, and computerized pilot avionics. CPS entails transdisciplinary approaches, merging concept of cybernetics, mechatronics, design and course of science. The method management is also known as embedded programs. In embedded programs, the emphasis tends to be extra on the computational parts, and fewer on an intense hyperlink between the computational and bodily parts. CPS can also be much like the Web of Issues (IoT), sharing the identical fundamental structure; however, CPS presents the next mixture and coordination between bodily and computational parts. Precursors of cyber-physical programs will be present in areas as various as aerospace, automotive, chemical processes, civil infrastructure, vitality, healthcare, manufacturing, transportation, leisure, and shopper home equipment.
Cyber-Physical Systems, a new formal paradigm to model redundancy and resiliency …
iPrescribe
On this paper, we current iPrescribe, a scalable low-latency structure for recommending ‘next-best-offers’ in a web-based setting. The paper presents the design of iPrescribe and compares its efficiency for implementations utilizing totally different real-time streaming expertise stacks. iPrescribe makes use of an ensemble of deep studying and machine studying algorithms for prediction. We describe the scalable real-time streaming expertise stack and optimized machine-learning implementations to attain a ninetieth percentile suggestion latency of 38 milliseconds. Optimizations embrace a novel mechanism to deploy recurrent Lengthy Quick Time period Reminiscence (LSTM) deep studying networks effectively. …
Concept Mask
Current works on semantic segmentation usually think about a small variety of labels, starting from tens to some a whole bunch. With a lot of labels, coaching and analysis of such process change into extraordinarily difficult as a consequence of correlation between labels and lack of datasets with full annotations. We formulate semantic segmentation as an issue of picture segmentation given a semantic idea, and suggest a novel system which may probably deal with an infinite variety of ideas, together with objects, components, stuff, and attributes. We obtain this utilizing a weakly and semi-supervised framework leveraging a number of datasets with totally different ranges of supervision. We first practice a deep neural community on a 6M inventory picture dataset with solely image-level labels to study visual-semantic embedding on 18K ideas. Then, we refine and lengthen the embedding community to foretell an consideration map, utilizing a curated dataset with bounding field annotations on 750 ideas. Lastly, we practice an attention-driven class agnostic segmentation community utilizing an 80-category absolutely annotated dataset. We carry out intensive experiments to validate that the proposed system performs competitively to the state-of-the-art on absolutely supervised ideas, and is able to producing correct segmentations for weakly realized and unseen ideas. …
