Quantum Computing: A Great Science in the Making
Andrew Chi-Chih Yao
In recent years, the scientific world has seen much excitement over the development of quantum computing, and the ever increasing possibility of building real quantum computers. What's the advantage of quantum computing? What are the secrets in the atoms that could potentially unleash such enormous power, to be used for computing and information processing? In this talk, we will take a look at quantum computing, and make the case that we are witnessing a great science in the making.
Free energy, the brain and life as we know it
How much about our interaction with – and experience of – our world can be deduced from basic principles? This talk reviews recent attempts to understand the self-organised behaviour of embodied agents, like ourselves, as satisfying basic imperatives for sustained exchanges with the environment. In brief, one simple driving force appears to explain many aspects of action and perception. This driving force is the minimisation of surprise or prediction error. In the context of perception, this corresponds to Bayes-optimal predictive coding that suppresses exteroceptive prediction errors. In the context of action, motor reflexes can be seen as suppressing proprioceptive prediction errors. We will look at some of the phenomena that emerge from this scheme, such as hierarchical message passing in the brain and the perceptual inference that ensues. I hope to illustrate these points using simple simulations of perception, action and action observation.
Modern rehabilitation methods in communication disorders
Communication is a complex human skill that combines physical and mental elements. Language and communication are crucial to all, particularly to children and young people. Modern society requires a high level of communication skills; speech, language, vision, and literacy are fundamental to meet these demands. Dysfunction in one or more of these areas, caused, for example, by various medical conditions, may lead to communication disorders that can involve impairment in hearing, vision and/or speech, which influences the ability to receive, comprehend, produce, and express verbal, nonverbal, and graphic information. Hearing impairment is one of the most common communication disorders, adversely affecting all aspects of life. The prevalence of hearing loss increases with age and may involve conductive hearing loss, acquired and delayed onset sensorineural hearing loss and/or auditory processing disorders that cannot be identified in neonatal hearing screening. Individuals with conductive hearing impairment can have either congenital or acquired middle ear pathology. Some forms of middle ear pathology, such as subtle middle ear abnormality with no apparent hearing loss and otitis media with effusion, are usually insidious and difficult for parents or care givers to detect. Sensorineural loss can result from inner ear defects or auditory nerve damage. Known causes of acquired sensorineural hearing loss include viral and bacterial infections, ototoxicity, and head trauma. Noise-related hearing impairment can also pose a serious health problem in children. Auditory processing disorders can result from neuromorphological disorders, maturational delay of the central auditory nervous system, and neurological disorders. Remarkable progress has been made during the last 3 decades in improving therapies for hearing impairment. Most of that progress is a result of technological and medical advances in cochlear implant technology and the use of digital technology for hearing aids. In the paper the state of the art medical technologies for treatment of hearing loss will be reviewed.
Ontologies on the Web: an alternative model
The Web has been a remarkable achievement for information sharing, and Linked Data is repeating doing the same for machine readable information. In my talk I will recapitulate the developments and motivation and underlying principles.
I will then revisit the notion of ontologies, which have been promoted and used for knowledge sharing. Several models for representing ontologies have been developed in the Knowledge Representation field, in particular associated with the Semantic Web.
I will argue that the currently advocated approaches miss certain basic properties of current distributed information sharing infrastructures (read: the Web and the Internet). I will present an alternative model aiming to support knowledge sharing and re-use on a global basis.
Natural Language as an Inspiration for a Unifying Framework for Computing
As a scientific discipline, the field of Computing lacks a unifying framework. It consists, instead, of diverse languages and techniques in the mostly disjoint areas of programming, databases, and artificial intelligence. However, the prospect of using formal logic to represent and reason with information extracted from the Web points the way to a unifying framework that is lacking in Computing today.
In my talk, I will propose a logic-based framework for Computing, which is inspired by natural language and by the distinction between goals and beliefs. I will argue that precisely written natural language texts can be viewed as programs to be executed by people, and are much higher-level than today’s programs that are designed to be executed by inanimate machines. In particular, the way that natural language seamlessly combines imperative and declarative sentences suggests a way in which the imperative features of programming languages and the declarative features of database systems and AI representations can also be combined. I will argue, moreover, that the relationship between imperative and declarative modes of expression corresponds to the relationship between goals and beliefs in the modelling of intelligent agents.
Theories of Clustering
Data clustering itself is an old subject but recently many more researchers study this issue, since its usefulness has been found in a variety of fields of informatics and applications. Although many techniques of clustering have been proposed, the theoretical basis is unknown to many researchers. Rather, clustering algorithms are regarded heuristic or ad hoc, with few exceptions. Such a view is partly true, since there are different origins of clustering. For example, numerical taxonomy uses hierarchical clusters. In contrast, pattern recognition regards K-means as a basic method that is quite different from hierarchical clustering. There is thus no unified theory that covers most of clustering algorithms. In other words, there are different theories that support different methods and algorithms. In this talk four topics that include two major theories:
- Transitive closure of weighted graphs related to agglomerative hierarchical clustering,
- Relation among K-means, fuzzy c-means, and mixture of distributions,
- Classifiers generated from different methods of clustering, and
- Positive-definite kernels applied to clustering algorithms.
This talk is mainly concerned with theoretical aspects of clustering, but illustrative examples are shown to help audience to intuitively understand essences of discussed methods.
Data Mining on Cloud Computing Platforms - Challenges and Solutions
Cloud computing has emerged rapidly as a growing paradigm of on-demand access to computing, data and software utilities using a usage-based billing model. Users essentially rent resources and pay for what they use and everything including software, platform, and infrastructure is as a service. Many massive data applications including data mining should be the ideal applications on cloud platforms. However, with the current cloud programming models, complicated data mining algorithms cannot be implemented easily and executed efficiently on the many cloud platforms. In this talk, I will give a review of different massively parallel computing platforms and compare various computing domains and programming models on these platforms, their limitations and potential solutions, especially to data mining applications. In particular, I will point out the shortcomings and limitations of current cloud computing programming models for typical data mining algorithms, and propose possible solutions. Current MapReduce model and its variants have succeeded in data-parallel applications such as database operations and web searching; however, they are still not effective for applications with a lot of data dependency such as data mining and graph applications. We propose several approaches to solving this problem through extension of current programming models, automatic translation from sequential codes to cloud codes, simple API and framework built on current cloud models, detection of data and task parallelism, and their efficient scheduling. Some preliminary theoretical and experimental results will also be reported in this talk.
John F. Sowa
In The Emperor’s New Mind, Roger Penrose claimed that quantum-mechanical effects are critical to human intelligence. But those effects need not be represented at the atomic level. A method of encoding conceptual graphs in continuous functions, which may be called quantum knowledge representation (QKR), exhibits the key properties of superposition, entanglement, and uncertainty. The operations of searching and graph matching, when performed on QKR, are analogous to the measurements in quantum-mechanical systems. A quantum computer would be ideal for processing them. But even with today’s digital computers, searching and graph matching on QKR can be performed with floating-point computations that scale in logarithmic time. For searching Big Data, they enable an ordinary laptop to outperform a supercomputer. This talk will show how QKR is used to analyze large volumes of documents and answer questions about them.
Light tools for tough jobs
Knowledge representation and reasoning are substantial ingredients of intelligent systems. The main challenges of knowledge representation manifest themselves in autonomous agent-based systems situated in dynamic and unpredictable environments. Traditional logical approaches to reasoning in such circumstances are usually too complex for real-world applications. As people are very effective in completing everyday tasks, modeling human-like mindset and rationality has been expected to substantially reduce the complexity of reasoning. In the course of development of knowledge representation techniques, many successful theoretical approaches have been proposed and investigated. However, the high expectations regarding modeling and reasoning are rarely met: theories frequently appear too complex both from computational and usability point of view.
The lecture will be devoted to a computationally friendly framework that makes modeling and reasoning intuitive for knowledge engineers. Surprisingly enough, a couple of simple ideas and constructs combined together appear sufficient to model heterogeneous information sources as well as to express advanced reasoning schemes handling incomplete, uncertain and inconsistent information. In particular, lightweight versions of a variety of reasoning methods, including nonmonotonic and defeasible ones can be expressed and combined arbitrarily in a well-understood and tractable manner. The key idea depends on structuring knowledge and applying carefully designed and easy to use rule-based calculus.
Last, but not least, the framework we advocate for is, among others, widely applicable to intelligent agent-based systems.