Search results for: supervised classifiers

Authors: Sofia Quaglia, Carmine Di Martino, Brendan Tierney

Abstract:

Artificial Intelligence (A.I.) constitutes various aspects of modern life, from the Machine Learning algorithms predicting the stocks on Wall streets to the killing of belligerents and innocents alike on the battlefield. Moreover, the end goal is to create autonomous A.I.; this means that the presence of humans in the decision-making process will be absent. The question comes naturally: when an A.I. does something wrong when its behavior is harmful to the community and its actions go against the law, which is to be held responsible? This research’s subject matter in A.I. and Robot Ethics focuses mainly on Robot Rights and its ultimate objective is to answer the questions: (i) What is the function of rights? (ii) Who is a right holder, what is personhood and the requirements needed to be a moral agent (therefore, accountable for responsibility)? (iii) Can an A.I. be a moral agent? (ontological requirements) and finally (iv) if it ought to be one (ethical implications). With the direction to answer this question, this research project was done via a collaboration between the School of Computer Science in the Technical University of Dublin that oversaw the technical aspects of this work, as well as the Department of Philosophy in the University of Milan, who supervised the philosophical framework and argumentation of the project. Firstly, it was found that all rights are positive and based on consensus; they change with time based on circumstances. Their function is to protect the social fabric and avoid dangerous situations. The same goes for the requirements considered necessary to be a moral agent: those are not absolute; in fact, they are constantly redesigned. Hence, the next logical step was to identify what requirements are regarded as fundamental in real-world judicial systems, comparing them to that of ones used in philosophy. Autonomy, free will, intentionality, consciousness and responsibility were identified as the requirements to be considered a moral agent. The work went on to build a symmetrical system between personhood and A.I. to enable the emergence of the ontological differences between the two. Each requirement is introduced, explained in the most relevant theories of contemporary philosophy, and observed in its manifestation in A.I. Finally, after completing the philosophical and technical analysis, conclusions were drawn. As underlined in the research questions, there are two issues regarding the assignment of moral agency to artificial agent: the first being that all the ontological requirements must be present and secondly being present or not, whether an A.I. ought to be considered as an artificial moral agent. From an ontological point of view, it is very hard to prove that an A.I. could be autonomous, free, intentional, conscious, and responsible. The philosophical accounts are often very theoretical and inconclusive, making it difficult to fully detect these requirements on an experimental level of demonstration. However, from an ethical point of view it makes sense to consider some A.I. as artificial moral agents, hence responsible for their own actions. When considering artificial agents as responsible, there can be applied already existing norms in our judicial system such as removing them from society, and re-educating them, in order to re-introduced them to society. This is in line with how the highest profile correctional facilities ought to work. Noticeably, this is a provisional conclusion and research must continue further. Nevertheless, the strength of the presented argument lies in its immediate applicability to real world scenarios. To refer to the aforementioned incidents, involving the murderer of innocents, when this thesis is applied it is possible to hold an A.I. accountable and responsible for its actions. This infers removing it from society by virtue of its un-usability, re-programming it and, only when properly functioning, re-introducing it successfully

Keywords: artificial agency, correctional system, ethics, natural agency, responsibility

Procedia PDF Downloads 156

Authors: Utkarsh A. Mishra, Ankit Bansal

Abstract:

At high temperature, radiative heat transfer is the dominant mode of heat transfer. It is governed by various phenomena such as photon emission, absorption, and scattering. The solution of the governing integrodifferential equation of radiative transfer is a complex process, more when the effect of participating medium and wavelength properties are taken into consideration. Although a generic formulation of such radiative transport problem can be modeled for a wide variety of problems with non-gray, non-diffusive surfaces, there is always a trade-off between simplicity and accuracy of the problem. Recently, solutions of complicated mathematical problems with statistical methods based on randomization of naturally occurring phenomena have gained significant importance. Photon bundles with discrete energy can be replicated with random numbers describing the emission, absorption, and scattering processes. Photon Monte Carlo (PMC) is a simple, yet powerful technique, to solve radiative transfer problems in complicated geometries with arbitrary participating medium. The method, on the one hand, increases the accuracy of estimation, and on the other hand, increases the computational cost. The participating media -generally a gas, such as CO₂, CO, and H₂O- present complex emission and absorption spectra. To model the emission/absorption accurately with random numbers requires a weighted sampling as different sections of the spectrum carries different importance. Importance sampling (IS) was implemented to sample random photon of arbitrary wavelength, and the sampled data provided unbiased training of MC estimators for better results. A better replacement to uniform random numbers is using deterministic, quasi-random sequences. Halton, Sobol, and Faure Low-Discrepancy Sequences are used in this study. They possess better space-filling performance than the uniform random number generator and gives rise to a low variance, stable Quasi-Monte Carlo (QMC) estimators with faster convergence. An optimal supervised learning scheme was further considered to reduce the computation costs of the PMC simulation. A one-dimensional plane-parallel slab problem with participating media was formulated. The history of some randomly sampled photon bundles is recorded to train an Artificial Neural Network (ANN), back-propagation model. The flux was calculated using the standard quasi PMC and was considered to be the training target. Results obtained with the proposed model for the one-dimensional problem are compared with the exact analytical and PMC model with the Line by Line (LBL) spectral model. The approximate variance obtained was around 3.14%. Results were analyzed with respect to time and the total flux in both cases. A significant reduction in variance as well a faster rate of convergence was observed in the case of the QMC method over the standard PMC method. However, the results obtained with the ANN method resulted in greater variance (around 25-28%) as compared to the other cases. There is a great scope of machine learning models to help in further reduction of computation cost once trained successfully. Multiple ways of selecting the input data as well as various architectures will be tried such that the concerned environment can be fully addressed to the ANN model. Better results can be achieved in this unexplored domain.

Keywords: radiative heat transfer, Monte Carlo Method, pseudo-random numbers, low discrepancy sequences, artificial neural networks

Procedia PDF Downloads 185

Authors: Hyohun Kim, Dongwha Shin, Yeonseok Kim, Ji-Su Ahn, Kensuke Nakamura, Dongeun Choi, Byung-Woo Hong

Abstract:

Artefacts are commonly encountered in the imaging process of clinical computed tomography (CT) where the artefact refers to any systematic discrepancy between the reconstructed observation and the true attenuation coefficient of the object. It is known that CT images are inherently more prone to artefacts due to its image formation process where a large number of independent detectors are involved, and they are assumed to yield consistent measurements. There are a number of different artefact types including noise, beam hardening, scatter, pseudo-enhancement, motion, helical, ring, and metal artefacts, which cause serious difficulties in reading images. Thus, it is desired to remove nuisance factors from the degraded image leaving the fundamental intrinsic information that can provide better interpretation of the anatomical and pathological characteristics. However, it is considered as a difficult task due to the high dimensionality and variability of data to be recovered, which naturally motivates the use of machine learning techniques. We propose an image restoration algorithm based on the deep neural network framework where the denoising auto-encoders are stacked building multiple layers. The denoising auto-encoder is a variant of a classical auto-encoder that takes an input data and maps it to a hidden representation through a deterministic mapping using a non-linear activation function. The latent representation is then mapped back into a reconstruction the size of which is the same as the size of the input data. The reconstruction error can be measured by the traditional squared error assuming the residual follows a normal distribution. In addition to the designed loss function, an effective regularization scheme using residual-driven dropout determined based on the gradient at each layer. The optimal weights are computed by the classical stochastic gradient descent algorithm combined with the back-propagation algorithm. In our algorithm, we initially decompose an input image into its intrinsic representation and the nuisance factors including artefacts based on the classical Total Variation problem that can be efficiently optimized by the convex optimization algorithm such as primal-dual method. The intrinsic forms of the input images are provided to the deep denosing auto-encoders with their original forms in the training phase. In the testing phase, a given image is first decomposed into the intrinsic form and then provided to the trained network to obtain its reconstruction. We apply our algorithm to the restoration of the corrupted CT images by the artefacts. It is shown that our algorithm improves the readability and enhances the anatomical and pathological properties of the object. The quantitative evaluation is performed in terms of the PSNR, and the qualitative evaluation provides significant improvement in reading images despite degrading artefacts. The experimental results indicate the potential of our algorithm as a prior solution to the image interpretation tasks in a variety of medical imaging applications. This work was supported by the MISP(Ministry of Science and ICT), Korea, under the National Program for Excellence in SW (20170001000011001) supervised by the IITP(Institute for Information and Communications Technology Promotion).

Keywords: auto-encoder neural network, CT image artefact, deep learning, intrinsic image representation, noise reduction, total variation

Procedia PDF Downloads 165

Authors: Tianqiao Zhang, Ye Tian, Yanliang Yin, Yichao Tian, Suzhai Tian, Weige Sun, Shuhui Gong, Limei Tang, Ruoliang Tang

Abstract:

Introduction: Healthcare workers, especially the nurses all over the world, are highly vulnerable to work-related musculoskeletal disorders (WMSDs), experiencing high rates of neck, shoulder, and low back injuries, due to the unfavorable working conditions. To reduce WMSDs among nursing personnel, many workplace interventions have been developed and implemented. Unfortunately, the ongoing Covid-19 (SARS-CoV-2) pandemic has posed great challenges to the ergonomic practices and interventions in healthcare facilities, particularly the hospitals, since current Covid-19 mitigation measures, such as social distancing and working remotely, has substantially minimized in-person gatherings and trainings. On the other hand, hospitals throughout the world have been short-staffed, resulting in disturbance of shift scheduling and more importantly, the increased job demand among the available caregivers, particularly the doctors and nurses. With the latest development in communication technology, remote intervention measures have been developed as an alternative, without the necessity of in-person meetings. The Omaha System (OS) is a standardized classification system for nursing practices, including a problem classification system, an intervention system, and an outcome evaluation system. This paper describes the development of an OS-based ergonomic intervention program. Methods: First, a comprehensive literature search was performed among worldwide electronic databases, including PubMed, Web of Science, Cochrane Library, China National Knowledge Infrastructure (CNKI), between journal inception to May 2020, resulting in a total of 1,418 scientific articles. After two independent screening processes, the final knowledge pool included eleven randomized controlled trial studies to develop the draft of the intervention program with Omaha intervention subsystem as the framework. After the determination of sample size needed for statistical power and the potential loss to follow-up, a total of 94 nurses from eight clinical departments agreed to provide written, informed consent to participate in the study, which were subsequently assigned into two random groups (i.e., intervention vs. control). A subgroup of twelve nurses were randomly selected to participate in a semi-structured interview, during which their general understanding and awareness of musculoskeletal disorders and potential interventions was assessed. Then, the first draft was modified to reflect the findings from these interviews. Meanwhile, the tentative program schedule was also assessed. Next, two rounds of consultation were conducted among experts in nursing management, occupational health, psychology, and rehabilitation, to further adjust and finalize the intervention program. The control group had access to all the information and exercise modules at baseline, while an interdisciplinary research team was formed and supervised the implementation of the on-line intervention program through multiple social media groups. Outcome measures of this comparative study included biomechanical load assessed by the Quick Exposure Check and stresses due to awkward body postures. Results and Discussion: Modification to the draft included (1) supplementing traditional Chinese medicine practices, (2) adding the use of assistive patient handling equipment, and (3) revising the on-line training method. Information module should be once a week, lasting about 20 to 30 minutes, for a total of 6 weeks, while the exercise module should be 5 times a week, each lasting about 15 to 20 minutes, for a total of 6 weeks.

Keywords: ergonomic interventions, musculoskeletal disorders (MSDs), omaha system, nurses, Covid-19

Procedia PDF Downloads 131

Authors: Oluwaseun Adeleke, Samuel O. Ikani, Simeon Christian Chukwu, Fidelis Edet, Anthony Nwala, Mopelola Raji, Simeon Christian Chukwu

Abstract:

Background: Health service providers are offered in-service training periodically to strengthen their ability to deliver services that are ethical, quality, timely and safe. Not all capacity-building courses have successfully resulted in intended service delivery outcomes because of poor training content, design, approach, and ambiance. The Delivering Innovations in Selfcare (DISC) project developed a Moment of Truth innovation, which is a proven training model focused on improving consumer/provider interaction that leads to an increase in the voluntary uptake of subcutaneous depot medroxyprogesterone acetate (DMPA-SC) self-injection among women who opt for injectable contraception. Methodology: Six months after training on a moment of truth (MoT) training manual, the project conducted two intensive rounds of qualitative data collection and triangulation that included provider, client, and community mobilizer interviews, facility observations, and routine program data collection. Respondents were sampled according to a convenience sampling approach, and data collected was analyzed using a codebook and Atlas-TI. Providers and clients were interviewed to understand their experience, perspective, attitude, and awareness about the DMPA-SC self-inject. Data were collected from 12 health facilities in three states – eight directly trained and four cascades trained. The research team members came together for a participatory analysis workshop to explore and interpret emergent themes. Findings: Quality-of-service delivery and performance outcomes were observed to be significantly better in facilities whose providers were trained directly trained by the DISC project than in sites that received indirect training through master trainers. Facilities that were directly trained recorded SI proportions that were twice more than in cascade-trained sites. Direct training comprised of full-day and standalone didactic and interactive sessions constructed to evoke commitment, passion and conviction as well as eliminate provider bias and misconceptions in providers by utilizing human interest stories and values clarification exercises. Sessions also created compelling arguments using evidence and national guidelines. The training also prioritized demonstration sessions, utilized job aids, particularly videos, strengthened empathetic counseling – allaying client fears and concerns about SI, trained on positioning self-inject first and side effects management. Role plays and practicum was particularly useful to enable providers to retain and internalize new knowledge. These sessions provided experiential learning and the opportunity to apply one's expertise in a supervised environment where supportive feedback is provided in real-time. Cascade Training was often a shorter and abridged form of MoT training that leveraged existing training already planned by master trainers. This training was held over a four-hour period and was less emotive, focusing more on foundational DMPA-SC knowledge such as a reorientation to DMPA-SC, comparison of DMPA-SC variants, counseling framework and skills, data reporting and commodity tracking/requisition – no facility practicums. Training on self-injection was not as robust, presumably because they were not directed at methods in the contraceptive mix that align with state/organizational sponsored objectives – in this instance, fostering LARC services. Conclusion: To achieve better performance outcomes, consideration should be given to providing training that prioritizes practice-based and emotive content. Furthermore, a firm understanding and conviction about the value training offers improve motivation and commitment to accomplish and surpass service-related performance outcomes.

Keywords: training, performance outcomes, innovation, family planning, contraception, DMPA-SC, self-care, self-injection.

Procedia PDF Downloads 50