The Pentus Theorem for Lambek Calculus with Simple Nonlogical Axioms

The Lambek calculus introduced in Lambek [6] is a strengthening of the type reduction calculus of Ajdukiewicz [1]. We study Associative Lambek Calculus L in Gentzen style axiomatization enriched with a finite set Γ of nonlogical axioms, denoted by L(Γ).It is known that finite axiomatic extensions of Associative Lambek Calculus generate all recursively enumerable languages (see Buszkowski [2]). Then we confine nonlogical axioms to sequents of the form p → q, where p and q are atomic types. For calculus L(Γ) we prove interpolation lemma (modifying the Roorda proof for L [10]) and the binary reduction lemma (using the Pentus method [9] with modification from [3]). In consequence we obtain the weak equivalence of the Context-Free Grammars and grammars based on L(Γ).     

Using artificial life to teach evolutionary biology

This paper presents a pilot study on the use of artificial life software in an educational setting. Two groups of high school students received a standard lesson in evolutionary biology followed by a software session. The experimental group used the suite of artificial life software presented in this paper; the control group used a commercial multimedia hypertext. At the end of the software session both groups were asked to fill in a simple multiple-choice questionnaire testing the students knowledge of various aspects of evolutionary biology. The results show that the group using the artificial life software performed significantly better than the control group. We argue that the experimental group may learn more effectively because the artificial life makes it possible for students to perform experiments, a possibility not available to the control group.

Testing for social learning and imitation in common marmosets, <Emphasis Type="Italic">Callithrix jacchus</Emphasis>, using an artificial fruit

We tested for social learning and imitation in common marmosets using an artificial foraging task and trained conspecific demonstrators. We trained a demonstrator marmoset to open an artificial fruit, providing a full demonstration of the task to be learned. Another marmoset provided a partial demonstration, controlling for stimulus enhancement effects, by eating food from the outside of the apparatus. We thus compared three observer groups, each consisting of four animals: those that received the full demonstration, those that received the partial demonstration, and a control group that saw no demonstration prior to testing. Although none of the observer marmosets succeeded in opening the artificial fruit during the test periods, there were clear effects of demonstration type. Those that saw the full demonstration manipulated the apparatus more overall, whereas those from the control group manipulated it the least of the three groups. Those from the full-demonstration group also contacted the particular parts of the artificial fruit that they had seen touched (localised stimulus enhancement) to a greater extent than the other two groups. There was also an interaction between the number of hand and mouth touches made to the artificial fruit for the full- and partial-demonstration groups. Whether or not these data represent evidence for imitation is discussed. We also propose that the clear differences between the groups suggest that social learning mechanisms provide real benefits to these animals in terms of developing novel food-processing skills analogous to the one presented here.     

Enhanced Verbal Statistical Learning in Glossolalia

Glossolalia (“speaking in tongues”) is a rhythmic utterance of word-like strings of sounds, regularly occurring in religious mass gatherings or various forms of private religious practices (e.g., prayer and meditation). Although specific verbal learning capacities may characterize glossolalists, empirical evidence is lacking. We administered three statistical learning tasks (artificial grammar, phoneme sequence, and visual-response sequence) to 30 glossolalists and 30 matched control volunteers. In artificial grammar, participants decide whether pseudowords and sentences follow previously acquired implicit rules or not. In sequence learning, they gradually draw out rules from repeating regularities in sequences of speech sounds or motor responses. Results revealed enhanced artificial grammar and phoneme sequence learning performances in glossolalists compared to control volunteers. There were significant positive correlations between daily glossolalia activity and artificial grammar learning. These results indicate that glossolalists exhibit enhanced abilities to extract the statistical regularities of verbal information, which may be related to their unusual language abilities.     

Surface features can deeply affect artificial grammar learning

Three experiments explored the extent to which surface features explain discrimination between grammatical and non-grammatical strings in artificial grammar learning (AGL). Experiment 1 replicated Knowlton and Squire’s (1996) paradigm using either letter strings as in the original study, or an analogous set of color strings to further explore if learning was affected by type of stimuli. Learning arose only with letter strings, but the results were mostly due to the discrimination of non-grammatical strings containing highly salient illegal features. Experiments 2 and 3 tested a new grammar devised to control for those features. Experiment 2 showed reduced grammar learning effects, and again only for letter materials. Experiment 3 explored the effect of additional practice with letter stimuli, and found increased learning only in the spaced practice condition, though additional practice also produced more explicit knowledge. These findings call for further research on the boundary conditions of learning in AGL paradigms.     

PSO-GA based hybrid with Adam Optimization for ANN training with application in Medical Diagnosis

This paper introduces a novel PSO-GA based hybrid training algorithm with Adam Optimization and contrasts performance with the generic Gradient Descent based Backpropagation algorithm with Adam Optimization for training Artificial Neural Networks. We aim to overcome the shortcomings of the traditional algorithm, such as slower convergence rate and frequent convergence to local minima, by employing the characteristics of evolutionary algorithms. PSO has a property of faster convergence rate, which can be exploited to account for the slower pace of convergence of the traditional BP (which is due to low values of gradients). In contrast, the integration with GA complements the drawback of convergence to local minima as GA, possesses the capability of efficient global search. So by this integration of these algorithms, we propose our new hybrid algorithm for training ANNs. We compare both the algorithms for the application of medical diagnosis. Results display that the proposed hybrid training algorithm, significantly outperforms the traditional training algorithm, by enhancing the accuracies of the ANNs with an increase of 20% in the average testing accuracy and 0.7% increase in the best testing accuracy.     

Mechanistic versus phenomenal embodiment: Can robot embodiment lead to strong AI?

Embodiment has become the raison d’etre for much of the new ‘cognitive robotics’. It fills a gap in the non-interactivist approach of traditional artificial intelligence (AI) in which ‘intelligence’ is viewed as the manipulation of symbols in a vacuum. However, a foundational question for the new AI is, can embodiment lead to a strong AI, i.e. a robot mind? To address this question, two extreme poles of embodiment are distinguished here, mechanistic and phenomenal. A detailed exploration of each type of embodiment is provided together with an appraisal of whether strong embodiment is possible for robotics, or whether robotics merely provides a tool for scientific exploration and modelling, i.e. weak embodiment? It is argued that strong embodiment, either mechanistic or phenomenal, is not possible for present day robots. However, weak embodiment may provide an enlightened approach to using robots for modelling cognition.     

On Learnability of Restricted Classes of Categorial Grammars

In this paper we present learning algorithms for classes of categorial grammars restricted by negative constraints. We modify learning functions of Kanazawa [10] and apply them to these classes of grammars. We also prove the learnability of intersection of the class of minimal grammars with the class of k-valued grammars. Presented by Wojciech Buszkowski     

Artificial organisms as tools for the development of psychological theory: Tolman’s lesson

In the 1930s and 1940s, Edward Tolman developed a psychological theory of spatial orientation in rats and humans. He expressed his theory as an automaton (the "schematic sowbug") or what today we would call an "artificial organism." With the technology of the day, he could not implement his model. Nonetheless, he used it to develop empirical predictions which tested with animals in the laboratory. This way of proceeding was in line with scientific practice dating back to Galileo. The way psychologists use artificial organisms in their work today breaks with this tradition. Modern "artificial organisms" are constructed a posteriori, working from experimental or ethological observations. As a result, researchers can use them to confirm a theoretical model or to simulate its operation. But they make no contribution to the actual building of models. In this paper, we try to return to Tolman's original strategy: implementing his theory of "vicarious trial and error" in a simulated robot, forecasting the robot's behavior and conducting experiments that verify or falsify these predictions.     

语义问题与人工智能模型构造的系统观点

基于"物理符号系统假设"的传统人工智能采信了低阶结构不连续的思想,将概念化与概念的语义基础分离,把对思维过程的模拟看作是能用形式化方法来实现的。但事实证明,这种后验性方法存在理论和实现上的双重危机,要完成有实际意义的、有创新性的智能行为丰富的语义是必需的,这使语义问题成为人工智能不同应用分支中的焦点,它包括语义获取、表达和使用三个方面。而要实现对语义问题的认识和解决就必须把它和其它智能行为作为一个连续的、相关的、不可分割的认知结构进行完整地考察,以系统的观点来看待智能模型的构造问题。这个认知结构统一性的基石就是基于神经生理基础的、以知觉的心理生理学解释为依据的、对语义的直接表达,这为统一以神经系统动力学为模型的其它各种智能行为提供了基础。