ANN-MIND : dropout for neural network training with missing data
- Authors: Mudau, Tshilidzi
- Date: 2019
- Subjects: Neural networks (Computer science) , Missing observations (Statistics) , Computer science
- Language: English
- Type: Masters (Thesis)
- Identifier: http://hdl.handle.net/10210/403506 , uj:33817
- Description: M.Sc. (Computer Science) , Abstract: It is a well-known fact that the quality of the dataset plays a central role in the results and conclusions drawn from the analysis of such a dataset. As the saying goes, ”garbage in, garbage out”. In recent years, neural networks have displayed good performance in solving a diverse number of problems. Unfortunately, neural networks are not immune to this misfortune presented by missing values. Furthermore, in most real-world settings, it is often the case that, the only data available for training neural networks consists of missing values. In such cases, we are left with little choice but to use this data for the purposes of training neural networks, although doing so may result in a poorly trained neural network. Most systems currently in use- merely discard the missing observation from the training datasets, while others just proceed to use this data and ignore the problems presented by the missing values. Still other approaches choose to impute these missing values with fixed constants such as means and mode. Most neural network models work under the assumption that the supplied data contains no missing values. This dissertation explores a method for training neural networks in the event where the training dataset consists of missing values...
- Full Text:
BioVault : a protocol to prevent replay in biometric systems
Digital environment evolution modelling and simulation
- Authors: Bengis, Merrick Kenna
- Date: 2020
- Subjects: Computer science , Computer simulation , Information technology
- Language: English
- Type: Doctoral (Thesis)
- Identifier: http://hdl.handle.net/10210/458387 , uj:40713
- Description: Ph.D. (Computer Science and Software Engineering) , Abstract: The concurrent growth of the human population and advancement in technology, together with ever-changing social interaction, has led to the creation of a large, abstract and complex entity known as the Digital Environment. In the current world, the Digital Environment, which is continually growing and ever-evolving, is now almost unrecognisable from what it started off as nearly 50 years ago. The human population has grown rapidly in the past century, growing to nearly 8 billion people in 2019, already double the population from 1975. This has created a world with more people than ever before, all of whom have a need to communicate with others, share information and form communities. Technology also experienced unprecedented advancements in this time, with important inventions such as electricity, computational machines, and communication networks. These technologies grew and allowed for people around the world to communicate as if they were next to each other, facilitated by the advent of the Internet. Presently, people all around the world are creating, sharing, and consuming information, while forming online communities, and also growing the physical footprint of the Internet and all connected devices. The intersection of these events formed the Digital Environment: an amalgamation of the physical, digital and cyber worlds. It is evident how rapidly and completely the Digital Environment has evolved in the past few decades, so what is in store for the future? Can people prepare for what the Digital Environment is to become and possibly even change its course? This thesis proposes a novel model for the simulation and prediction of the evolution of the Digital Environment: the Digital Environment Evolution Modelling and Simulation model or DEEv-MoS. The DEEv-MoS model proposes a method that makes use of well-developed and commonly used fields of research to create a holistic simulation of the Digital Environment and its many parts. Through the use of intelligent agents, entity component systems and machine learning, accurate simulations can be run to determine how the future digital landscape will grow and change. This allows researchers to further understand what the future holds and prepare for any eventualities, whether they are positive or negative...
- Full Text:
Keuring en opleiding van rekenaarstelselontleders.
- Authors: Raubenheimer, I. van W. , Fick, L.J.
- Date: 1980
- Subjects: Selection of computer systems , Training of computer systems , Computer systems analysts , Job description , Computer science
- Type: Article
- Identifier: uj:6542 , http://hdl.handle.net/10210/2741
- Description: Some of the results of an extensive study on selection and training of computer systems analysts are reported. Special attention is devoted to a job description and job analysis as a basis for identifying the critical attributes and training requirements involved. The development and validation of a battery for the selection of computer systems analysts and students of computer science are discussed.
- Full Text:
Wat is rekenaarwetenskap?
- Authors: Smith, T. H. C.
- Date: 2009-05-07T07:12:34Z
- Subjects: Computer science - Study and teaching (Higher) - South Africa , Computer science
- Type: Inaugural
- Identifier: uj:15006 , http://hdl.handle.net/10210/2488
- Description: Inaugural lecture--Department of Computer Science, Rand Afrikaans University, 20 May 1986 , Computer Science as discipline is involved with the study of computers and the phenomena connected with computing, notably algorithms: programs and programming. In the lecture this definition is elucidated by discussing computers and algorithms in greater detail. Thereafter attention is drawn to the relationship between Computer Science and other disciplines. Finally undergraduate curricula in Computer Science are discussed.
- Full Text: