On connectionism, rule extraction, and brain-like learning

There is a growing body of work that shows that both fuzzy and symbolic rule systems can be implemented using neural networks. This body of work also shows that these fuzzy and symbolic rules can be retrieved from these networks, once they have been learned by procedures that generally fall under the category of rule extraction. The paper argues that the idea of rule extraction from a neural network involves certain procedures, specifically the reading of parameters from a network, that are not allowed by the connectionist framework that these neural networks are based on. It argues that such rule extraction procedures imply a greater freedom and latitude about the internal mechanisms of the brain than is permitted by connectionism, but that such latitude is permitted by the recently proposed control theoretic paradigm for the brain. The control theoretic paradigm basically suggests that there are parts of the brain that control other parts and has far less restrictions on the kind of procedures that can be called “brain like”. The paper shows that this control theoretic paradigm is supported by new evidence from neuroscience about the role of neuromodulators and neurotransmitters in the brain. In addition, it shows that the control theoretic paradigm is also used in connectionist algorithms, although never acknowledged explicitly. The paper suggests that far better learning and rule extraction algorithms can be developed using these control theoretic notions and they would be consistent with the more recent understanding of how the brain works and learns     

On the Fly Estimation of the Processes that Are Alive in an Asynchronous Message-Passing System

It is well known that in an asynchronous system where processes are prone to crash, it is impossible to design a protocol that provides each process with the set of processes that are currently alive. Basically, this comes from the fact that it is impossible to distinguish a crashed process from a process that is very slow or with which communications are very slow. Nevertheless, designing protocols that provide the processes with good approximations of the set of processes that are currently alive remains a real challenge in fault-tolerant-distributed computing. This paper proposes such a protocol, plus a second protocol that allows to cope with heterogeneous communication networks. These protocols consider a realistic computation model where the processes are provided with nonsynchronized local clocks and a function alpha () that takes a local duration Delta as a parameter, and returns an integer that is an estimate of the number of processes that could have crashed during that duration Delta. A simulation-based experimental evaluation of the proposed protocols is also presented. These experiments show that the protocols are practically relevant.     

Gödel,Penrose and the possibility of AI

It is argued that Penrose's use of Gödel's theorem to show that no machine can reproduce the achievements of a human mathematician depends on extra assumptions that would also show that no machine could ever check the validity of mathematical proofs; and it is claimed that the latter result contradicts views about the possibility of communicating mathematical proofs that are widely held and that are endorsed by Penrose.     

Searle��s wager

Nicholas Agar has recently argued that it would be irrational for future human beings to choose to radically enhance themselves by uploading their minds onto computers. Utilizing Searle’s argument that machines cannot think, he claims that uploading might entail death. He grants that Searle’s argument is controversial, but he claims, so long as there is a non-zero probability that uploading entails death, uploading is irrational. I argue that Agar’s argument, like Pascal’s wager on which it is modelled, fails, because the principle that we (or future agents) ought to avoid actions that might entail death is not action guiding. Too many actions fall under its scope for the principle to be plausible. I also argue that the probability that uploading entails death is likely to be lower than Agar recognizes.     

Vulnerability analysis of immunity-based intrusion detection systems using genetic and evolutionary hackers

Artificial immune systems (AISs) are biologically inspired problem solvers that have been used successfully as intrusion detection systems (IDSs). In this paper, we compare a genetic hacker with 12 evolutionary hackers based on particle swarm optimization (PSO) that have been effectively used as vulnerability analyzers (red teams) for AIS-based IDSs. Our results show that the PSO-based red teams that use Clerc's constriction coefficient outperform those that do not. Our results also show that the three types of red teams (genetic, basic PSO, and PSO with the constriction coefficient) have distinct search behaviors that are complimentary.     

Decidable metric logics

The common metric temporal logic for continuous time were shown to be insufficient, when it was proved that they cannot express a modality suggested by Pnueli. Moreover no finite temporal logic can express all the natural generalizations of this modality. It followed that if we look for an optimal decidable metric logic we must accept infinitely many modalities, or adopt a different formalism.Here we identify a fragment of the second order monadic logic of order with the “+1” function, that expresses all the Pnueli modalities and much more. Its main advantage over the temporal logics is that it enables us to say not just that within prescribed time there is a point where some punctual event will occur, but also that within prescribed time some process that starts now (or that started before, or that will start soon) will terminate. We prove that this logic is decidable with respect to satisfiability and validity, over continuous time. The proof depends heavily on the theory of compositionality. In particular every temporal logic that has truth tables in this logic is automatically decidable. We extend this result by proving that any temporal logic, that has all its modalities defined by means more general than truth tables, in a logic stronger than the one just described, has a decidable satisfiability problem. We suggest that this monadic logic can be the framework in which temporal logics can be safely defined, with the guarantee that their satisfiability problem is decidable.     

Practical dynamic consensus of Stuart–Landau oscillators over heterogeneous networks

ABSTRACT

In Panteley and Loria (2017), a framework for the study of synchronisation and collective behaviour of networked heterogeneous systems was introduced. It was underlined that in such scenario an emergent collective behaviour arises, one that is inherent to the network and that is independent of the interconnection strength. Therefore, the natural way to make complete study of synchronisation is by investigating, on one hand, the stability of the emergent dynamical system and, on the other, by assessing the difference between the motion of each individual system and that of the emergent one. Thus, if all systems' motions approach that of the emergent dynamics, we say that they reach dynamic consensus. In this paper, we study dynamic consensus of a fairly general class of nonlinear heterogeneous oscillators, called Stuart–Landau. We establish that the emergent dynamics consists in that of an ‘averaged’ oscillator with a global attractor that consists in a limit-cycle and, moreover, we determine its frequency of oscillation. Then, we show that the heterogeneous oscillators achieve practical dynamic consensus, that is, their synchronisation errors measured relative to the collective motion are ultimately bounded.     

A computational framework for concept formation for a situated design agent

This paper takes the approach that designing is situated and that concepts are formed as a consequence of the situatedness of designing. The paper presents a framework for concept formation that draws on a structure of a design agent that includes sensors, perceptors and conceptors that interact with each other and the external and internal environment of the agent to produce the situation that is a contingent basis for the formation and use of concepts.     

Moving the goal posts: a reply to Dawson and Piercey

Berkeley [Minds Machines 10 (2000) 1] described a methodology that showed the subsymbolic nature of an artificial neural network system that had been trained on a logic problem, originally described by Bechtel and Abrahamsen [Connectionism and the mind. Blackwells, Cambridge, MA, 1991]. It was also claimed in the conclusion of this paper that the evidence was suggestive that the network might, in fact, count as a symbolic system. Dawson and Piercey [Minds Machines 11 (2001) 197] took issue with this latter claim. They described some lesioning studies that they argued showed that Berkeley’s (2000) conclusions were premature. In this paper, these lesioning studies are replicated and it is shown that the effects that Dawson and Piercey rely upon for their argument are merely an artifact of a threshold function they chose to employ. When a threshold function much closer to that deployed in the original studies is used, the significant effects disappear.     

Understanding compulsive social media use: The premise of complementing self-conceptions mismatch with technology

Some anecdotal accounts and research reports have suggested that obsessive social media involvement could turn into a compulsive behavior among university students. Unfortunately, the research that sheds light on the possible conditional nature of that relationship is scarce at best. Therefore, this study tries to address this issue by developing a contingency-based model and tests it using data gathered from a sample of university students. The model postulates that compulsive social media use arises due to self-awareness factors, and together they in turn predict problematic learning outcomes. It also postulates that these relationships are moderated by the influence of technological factors. The results indicate that self-esteem has a significant negative influence on compulsive social media use and that interaction anxiousness has a significant positive influence on the same. The results also reveal that only compulsive social media use has a significant direct influence on problematic learning outcomes; and that social media complementarity plays a moderating role in the model. We discuss the implications of these findings for research and practice.     

Subtleties of Transactional Memory Atomicity Semantics

Transactional memory has great potential for simplifying multithreaded programming by allowing programmers to specify regions of the program that must appear to execute atomically. Transactional memory implementations then optimistically execute these transactions concurrently to obtain high performance. This work shows that the same atomic guarantees that give transactions their power also have unexpected and potentially serious negative effects on programs that were written assuming narrower scopes of atomicity. We make four contributions: (1) we show that a direct translation of lock-based critical sections into transactions can introduce deadlock into otherwise correct programs, (2) we introduce the terms strong atomicity and weak atomicity to describe the interaction of transactional and non-transactional code, (3) we show that code that is correct under weak atomicity can deadlock under strong atomicity, and (4) we demonstrate that sequentially composing transactional code can also introduce deadlocks. These observations invalidate the intuition that transactions are strictly safer than lock-based critical sections, that strong atomicity is strictly safer than weak atomicity, and that transactions are always composable     

A new paradigm for distributed,integrated, real-time control systems

To exploit the potential offered by integrated intelligent control systems, it is essential that underlying architectures are available that are able to integrate not only hardware and software, but also human beings into the control loop. By studying the way that a human workforce operates, this paper suggests how a human systems analogy can be drawn up that highlighted some fundamental differences between the way that humanbased, and computer-based systems operate. One of the key differences is that humans are able to reason not only logically, but also in terms of time. This paper addresses the philosphy behind the DENIS architecture — a distributed architecture that enables this high level of compatability between autonomous intelligent agents.     

The Relaxed Online Maximum Margin Algorithm

We describe a new incremental algorithm for training linear threshold functions: the Relaxed Online Maximum Margin Algorithm, or ROMMA. ROMMA can be viewed as an approximation to the algorithm that repeatedly chooses the hyperplane that classifies previously seen examples correctly with the maximum margin. It is known that such a maximum-margin hypothesis can be computed by minimizing the length of the weight vector subject to a number of linear constraints. ROMMA works by maintaining a relatively simple relaxation of these constraints that can be efficiently updated. We prove a mistake bound for ROMMA that is the same as that proved for the perceptron algorithm. Our analysis implies that the maximum-margin algorithm also satisfies this mistake bound; this is the first worst-case performance guarantee for this algorithm. We describe some experiments using ROMMA and a variant that updates its hypothesis more aggressively as batch algorithms to recognize handwritten digits. The computational complexity and simplicity of these algorithms is similar to that of perceptron algorithm, but their generalization is much better. We show that a batch algorithm based on aggressive ROMMA converges to the fixed threshold SVM hypothesis.     

Simple evolution of complex crystal species

Cairns-Smith has proposed that life began as structural patterns in clays that self-replicated during cycles of crystal growth and fragmentation. Complex, evolved crystal forms could then have catalyzed the formation of a more advanced genetic material. A crucial weakness of this theory is that it is unclear how complex crystals might arise through Darwinian evolution and selection. Here we investigate whether complex crystal patterns could evolve using a model system for crystal growth, DNA tile crystals, that is amenable to both theoretical and experimental inquiry. It was previously shown that in principle, the evolution of crystals assembled from a set of thousands of DNA tile types under very specific environmental conditions could produce arbitrarily complex patterns. Here we show that evolution driven only by the dearth of one monomer type could produce complex crystals from just 12 monomer types. When a monomer type is rare, crystals that use few of this monomer type are selected for. We use explicit enumeration to show that there are situations in which crystal species that use a particular monomer type less frequently will grow faster, yet to do so requires that the information contained in the crystal become more complex. We show that this feature of crystal organization could allow more complex crystal morphologies to be selected for in the right environment, using both analysis in a simple model of self-assembly and stochastic kinetic simulations of crystal growth. The proposed mechanism of evolution is simple enough to test experimentally and is sufficiently general that it may apply to other DNA tile crystals or even to natural crystals, suggesting that complex crystals could evolve from simple starting materials because of relative differences in concentrations of the materials needed for growth.     

Symmetry between the intentionality of minds and machines? The biological plausibility of Dennett’s account

One of the most influential arguments against the claim that computers can think is that while our intentionality is intrinsic, that of computers is derived: it is parasitic on the intentionality of the programmer who designed the computer-program. Daniel Dennett chose a surprising strategy for arguing against this asymmetry: instead of denying that the intentionality of computers is derived, he endeavours to argue that human intentionality is derived too. I intend to examine that biological plausibility of Dennett’s suggestion and show that Dennett’s argument for the claim that human intentionality is derived because it was designed by natural selection is based on the misunderstanding of how natural selection works.     

Tone mapping to minimize visual sensation distortion considering brightness and local band‐limited contrast

We propose a tone‐mapping algorithm that minimizes the function that represents the visual sensation distortion that occurs after tone mapping. For the function, we consider both brightness and local band‐limited contrast to reduce darkening of images that have high dynamic range when they are displayed with conventional contrast‐based tone mapping using histogram on devices that have low dynamic range. By exploiting human visual characteristics, we simplify the problem and find a closed‐form solution that minimizes the function that represents distortion of visual sensation. In both subjective and objective evaluations, the proposed algorithm achieves a processed image that is most similar to the original image and has the best subjective image quality.     

Digital forensics and the issues of identity

The issue of what we consider to be the identity of a person has become increasingly complex as we have made ever greater use of the facilities and services that have been made available by developing technologies and the Internet. In the past people normally had one identity, while in the current environment it is acceptable to maintain separate ‘identities’ for different aspects of our on-line interactions.Proving beyond a reasonable doubt that an individual that is suspected of a crime that is based on the technologies that we increasingly rely on was the actual perpetrator has always been problematic. It is relatively easy to determine the device that was used, but proving that the suspect was the person that used it has always been more difficult.This paper looks at a range of issues that have affected what we consider to be reasonable proof of identity and a number of the problems that this causes in identifying the perpetrator of a crime.