There are three model stages in the workflow Atlas supports – experimentation (selection), eval- uation, and production (distribution). These have motivated a system that uniquely fulfills the non-functional requirements of usability, extensibility, and flexibility. While many organizations have departmental separation of these states – where data scientists develop prototype models and software engineers create structured implementations of them – a finite resource pool drove the need for a consistent interface between these stages. Unfortunately, these stages can often introduce requirements in conflict with one another. Atlas was conceived as a way to mitigate those competing requirements while allowing data scientists to write flexible experiments that could be more readily refined for production.
Medical predictive analytics have gained popularity in recent years, with numerous publications focusing on models that estimate patients’ risk of a disease or a future health state (the ‘event’) based on classical regression algorithms or modern flexiblemachine learning or artificial intelligence algorithms [1–3]. These predictions may support clinical decision- making and better inform patients. Algorithms (or risk prediction models) should give higher risk esti- mates for patients with the event than for patients without the event (‘discrimination’). Typically, dis- crimination is quantified using the area under the receiver operating characteristic curve (AUROC or AUC), also known as the concordance statistic or c- statistic. Additionally, it may be desirable to present classification performance at one or more risk thresholds such as sensitivity, specificity, and
We proposed a simple, powerful, and flexiblemachine learn- ing framework for (i) reducing the search space of computa- tionally difficult enumeration problems and (ii) augmenting existing state-of-the-art solvers with informative cues aris- ing from the input distribution. In particular, we focused on a probabilistic preprocessing strategy, which retained all maximum cliques on a representative selection of large real-world networks from different domains. We showed the practicality of our framework by showing it speeds up the execution of state-of-the-art solvers on large graphs without compromising the solution quality. In addition, we demon- strated that supervised learning is a viable approach for tack- ling NP-hard problems in practice.
We consider an ATM center with a fixed and a flexiblemachine. Customers arrive according to a Poisson process at a rate λ for service on any available machine present in the system upon arrival. We suppose that the fixed machine has a faster service rate compared with the flexible one. Thus, the fixed machine Received: September 18, 2016 c 2017 Academic Publications
(1994), Dunker et al (2005),Ming et al (2009) ,Junfeng et al(2011) for the dynamic layout problem. The major disadvantages of using these QAP based approaches are that shape constraints cannot be considered in the design stage and facility sites are known and have equal sizes. Hence, they are not well suited for the flexiblemachine layout problem discussed in this paper. However, some conceptual ideas from these approaches will be used to build a solution methodology for this problem. Montreuil and Laforge (1992) also solve a dynamic layout design problem given a scenario tree of probable future scenarios. Several layout design alternatives for the future are constructed based on this tree. They use a design skeleton approach to determine the machine layout. Lacksonen (1994) uses Montreuil’s formulation for a machine layout design problem to formulate a dynamic layout design problem. This formulation is solved with a two-step procedure similar to the idea of a design skeleton for the machine layout problem. Both Montreuil’s approach and Lacksonen’s approach allow the machines to have varying areas but not a fixed shape or geometry. Furthermore, the pickup and drop-off stations are allowed to move in order to further reduce the objective value.
This section is related to the application of artificial intelligence (AI) in the proposed theory. In recent years, the application of AI methods in engineering sciences is very common. Methods such as artificial neural networks (ANN) [34-39], radial basis function (RBF) [40-44], genetic programming (GP) [45-49], genetic algorithm (GA) [50-53], gene expression programming (GEP) [24, 54, 55], support vector machine (SVM) [40, 54, 56-58], Random Forest (RF) [59-63], Fuzzy systems [64-67], and regression tree (RT) [68-70] have received much attention from engineers . In this paper, authors use the RF and Random Forest optimized by Genetic Algorithm (RF-GA) methods to predict PCI based on IRI. These methods were introduced in the next sections.
Flexible manufacturing system has come up as a mean to achieve these prerequisites term flexible manufacturing system, or FMS, refers to a highly automated GT machine cell, consisting of a group of computer numerical control (CNC) machine tools and supporting workstations, interconnected by an automated material handling and storage system, and all controlled by a distributed computer system. The reason, the FMS is called as flexible. Due to the benefits of manufacturing systems, it is intended to make it flexible, durable, simplicity, and reliable. To be a flexible, the manufacturing system must possess the following capabilities such capabilities are often difficult to engineer through manual operations. So it is necessary to assist an automated system with sensor system which is required to accomplish wants of business environment. It has been taken several years together to make manufacturing systems flexible in all various classes which were explained in brief as follows:
Communication networks represent several important areas for study. These include graph theory, connectivity, algorithm design, protocol engineering and many others. The advent of the Internet of Things (IoT), 5G and the increased use of virtualization have created new, or least magnified, problems for modern communication networks. These problems include a very high number of de- vices, nodes that come and go with increased frequency, demands on capacity, quality of service/experience, security threats and visibility black holes. While there are many aspects to each, of critical importance is an understanding of data packets and packet flows. Today, the ability to apply more powerful algo- rithms and greater processing power has facilitated the application of machine learning to these network data streams. As a result, machine learning tech- niques and architectures to address network visibility, quality of service and security have become major focal points for both industry and academia. However, applying machine learning techniques to network problems is not without its own issues. Researchers experiment in the blind with structures or hyper-parameters without reasoning behind the choices made. It is common to see architectures utilized simply because it was successful for some other application. Thus, available works do not document their rationale for choices made or fail to experiment with varying parameters because they were not part of the problem studied. The central difficulty with this approach is that researchers lack a body of knowledge that can be used as a basis for the work.
 S. Dumais, J. Platt, D. Heckerman, and M. Sahami, “Inductive Learning Algorithms and Representations for Text Categorization,” Proc. Seventh Int’l Conf. Information and Knowledge Management (CIKM ’98), pp. 148-155, 1998.  D.D. Lewis, “An Evaluation of Phrasal and Clustered Representations on a Text Categorization Task,” Proc. 15th ACM Int’l Conf. Research and Development in Information Retrieval (SIGIR ’92), N.J. Belkin, P. Ingwersen, and A.M. Pejtersen, eds., pp. 37-50, 1992.  R.E. Schapire and Y. Singer, “Boostexter: A Boosting-Based System for Text Categorization,” Machine Learning, vol. 39, nos. 2/3, pp. 135-168, 2000.
The efficiency and quality verification of plastic shell circuit breaker screw mounting machine is accomplished by comparing the man-made installation and the machine installation.The efficiency of screw mounting is measured via the number of the installed circuit breakers,while the quality of the screw mounting is measured through the degree of deflection of the screw mounting and the preload.The experiment result shows that the efficiency of the screw mounting machine is 12 times higher than that of the manual operation, and the installation quality is significantly higher than the manual installation.
For Hologic Selenia Dimensions most of the calibration is done automatically. A 2 and 8 cm thick phantom (BR-12) is compressed by applying 133.5 N compression force, and the machine will then register the thickness of the phantom. For the “FAST” paddle (the flexible paddle) the same approach is taken, but without any compression. The paddle is just lowered until it touches the phantom, and the machine is told that this is 2 or 8 cm. The fact that a rigid phantom is used for this test is probably not optimal, because a tilt will probably occur. Maybe one needs to rethink how the thickness is measured, or maybe a different approach to how the paddle is constructed needs to be addressed.
lems for the objectives of minimizing the total comple- tion time and maximum lateness, respectively, where the number and the starting time of unavailability constraint are decision variables. They showed that these problems are NP-hard. Sbihi and Varnier  presented a heuristic method for the single-machine scheduling problem with several maintenance periods. Specically, two situations were investigated in their study: rst, maintenance periods were periodically xed (i.e. 1; h i jpaT max ); second, maintenance periods
Given that the utilization of machine learning (ML) for solving the nonlinear problems has become very popular in the past years, this paper first tries to fix the mentioned problem about the data analysis method in the prior paragraph. For this purpose, the authors used three ML techniques, including Gaussian process regression (GPR), m5p model tree (M5P) and random forest (RF), to estimate SN eff . The other problem mentioned in the former paragraph
Peel forces of participants compared to machine peel forces for different seal temperatures of each sealed film are shown in Fig. 4-6. Statistically significant differences (p<0.05) are displayed via different letters on the top of each bar in each figure. According to the results, average maximum machine peel forces of the seal for LDPE and LLDPE 70µm and Nylon15/LLDPE45 at 112ºC were 15.91, 9.36 and 10.40 N/15mm respectively. Although, LLDPE generally produces a stronger seal than LDPE due to the compact molecular structure, for this test, LDPE showed the highest seal strength at low seal temperature since its heat transfer rate is higher than LLDPE and multilayer Nylon/LLDPE. Therefore, higher heat transfer rate leads to a higher surface temperature and will result in a stronger seal for LDPE. When the seal temperature is increased, we can see that Nylon/LLDPE shows a stronger seal than LDPE because the sealing layer of LLDPE in the laminated structure received high heat and produced a strong molecular bonding.
ity constraint, Yang et.al  proved that the single machine scheduling problem with a flexible maintenance by minimizing the makespan is NP-hard. They pro- vided a heuristic algorithm with complexity O(n log n) to find a good solution for the problem. Chen  proposed two mixed binary integer programming (BIP) models for solving the single machine scheduling problem with a flexible mainte- nance to minimize the total tardiness (denoted by 1, h 1 |f| P T i ), while Chen 
Kim, M., Tutumluer, E., and Kwon, J. (2009) “Nonlinear pavement foundation modeling for three-dimensional finite-element analysis of flexible pavements”, International Journal of Geomechanics, Vol. 9, Number 5, pp. 195-208. Lin, J. and Liu, Y. (2010) “Potholes detection based on SVM in the pavement distress image”, Paper presented at the Ninth International Symposium on Distributed Computing and Applications to Business Engineering and Science (DCABES), Hong Kong, China.