This paper describes NUTRITION UCR, a prototype expert system for human nutritional diagnosis developed in Java on Android using a service-oriented architecture. The system runs on mobile devices and offers smart features that evaluate the nutritional condition of an individual by assessing their physical characteristics and eating habits. We explain the knowledge engineering process used to develop the system, overview the system architecture and selected design tools, and summarize some preliminary results from the prototype implementation.

The thalamus is a critical brain structure involved in gating and regulating the flow of sensory and cortical information. The reticular nucleus of the thalamus (TRN) sends inhibitory projections to the thalamic relay nuclei instead of projecting to the cortex as the other thalamic nuclei do. These inhibitory projections endow the TRN with the functionality to modulate and control cortical information flowing through the thalamus. Yet, the functional roles of the TRN and thalamus in high-level cognitive processing, such as spatial reasoning and decision-making, remains poorly understood. Neurocognitive models offer a framework to explore the high-level cognitive functions of the thalamus and TRN. Here, we investigate the functional roles of the thalamus and TRN in high-level cognitive tasks using a large-scale neurocognitive model called ICArUS-MINDS. Our results demonstrate distributed and parallel top-down executive control of semantic and spatial cortical information. Specifically, we observed reticular-thalamic inhibitory gating of spatial and semantic information through top-down task switching control during reasoning, decision making and recall. Thalamic-gating was critical for orchestrating processing sequences of task-dependent switches between cortical sources and targets. These results are an important first step in simulating and understanding the functional roles and behaviors of the thalamic brain system in high-level cognitive processing.

Many practical problems in multiple fields can be converted to a SAT problem, or a sequence of SAT problems, such that their solution immediately implies a solution to the original problem. Despite the enormous progress achieved over the last decade in the development of SAT solvers, there is strong demand for higher algorithm efficiency to solve harder and larger problems. The widespread availability of multi-core, shared memory parallel environments provides an opportu- nity for such improvements. In this paper we present our re- sults on improving the effectiveness of standard SAT solvers on such architectures, through a portfolio approach. Multiple instances of the same basic solver using different heuristic strategies, for search-space exploration and problem analy- sis, share information and cooperate towards the solution of a given problem. Results from the application of our methodol- ogy to known problems from SAT competitions show relevant improvements over the state of the art and yield the promise of further advances.

Physical or simulated agents sharing an environmentwith humans must evaluate the impact of their own and other agents'actions in the specific social and cultural context. It is desirablethat this social calculus aligns itself with the models developed insociology and psychology — however, it needs to be expressed in anoperational, algorithmic form, suitable for implementation. While we can develop the framework of social calculus based onpsychological theories of human behavior, the actual form of thealgorithms can only be acquired from the knowledge of the specificculture. In this paper we consider social calculus based onculture-sanctioned social values (CSSMs). A critical component of thismodel is the set of action-impact functions (AIFs), which describe howthe actions of the agents change the CSSMs in specific settings. Wedescribe a technique to evolve the AIFs using genetic programming basedon a limited set of data pairs which can be obtained by surveying humansimmersed in the specific culture. We describe the proposed model througha scenario involving a group of soldiers and a robot acting on apeacekeeping mission.

While the use of abstraction and its benefit in terms of transferring learned information to new tasks has been studied extensively and successfully in MDPs, it has not been studied in the context of Partially Observable MDPs. This paper addresses the problem of transferring skills from previous experiences in POMDP models using high-level actions (options). It shows that the optimal value function remains piecewise-linear and convex when policies are high-level actions, and shows how value iteration algorithms can be modified to support options. The results can be applied to all existing value Iteration algorithms. Experiments show how adding options can speed up the learning process.

This paper describes a method for automatically imitating a particular facial expression in an avatar through a hybrid Particle Swarm Optimization – Tabu Search algorithm. The muscular structures of the facial expressions are measured by Ekman and Friesen’s Facial Action Coding System (FACS). Using a neutral expression as a reference, the minute movements of the Action Units, used in FACS, are automatically tracked and mapped onto the avatar using a hybrid method. The hybrid algorithm is composed of Particle Swarm Optimization algorithm and Tabu Search. Distinguishable features portrayed on the avatar ensure a personalized, realistic imitation of the facial expressions. To evaluate the feasibility of using PSO-TS in this approach, a fundamental proof-of-concept test is employed on the system using the OGRE avatar. Results are described and discussed.

This study presents a new method for knowledge transfer in Multi-Population Cultural Algorithms (MPCA) through agent migration. This agent-based algorithm involves having individual agents using one of multiple pre-defined knowledge algorithms to de-termine behavior, and using the success of it and other agents to decide on which knowledge algorithms to use next. Two or more subpopulations with their own knowledge algorithm are created. The agents work in the same environment by only communicating with agents within their own subpopulation, and with two global belief spaces monitoring the effectiveness of each subpopulation. Agents transfer between the sub-populations regularly to further improve individual success. We use the "cone’s world" problem as test-bed. Experimental results reveal the impact of indi-vidual knowledge transfer on the target subpopula-tion’s belief space.

Author identification is a subfield of Natural Language Processing (NLP) that uses machine learning techniques to identify the author of a text. Most previous research focused on long texts with the assumption that a minimum text length threshold exists under which author identification would no longer be effective. This paper examines author identification in short texts far below this threshold, focusing on messages retrieved from Twitter (maximum length: 140 characters) to determine the most effective feature set for author identification. Both Bag-of-Words (BOW) and Style Marker feature sets were extracted and evaluated through a series of 15 experiments involving up to 12 authors with large and small dataset sizes. Support Vector Machines (SVM) were used for all experiments. Our results achieve classification accuracies approaching that of longer texts, even for small dataset sizes of 60 training instances per author. Style Marker feature sets were found to be significantly more useful than BOW feature sets as well as orders of magnitude faster, and are therefore suggested for potential applications in future research.

Lexical chains between two concepts are sequences of semantically related words interconnected via semantic relations. This paper presents a new approach for the automatic construction of lexical chains on knowledge bases. Experiments were performed building lexical chains on WordNet, Extended WordNet, and Extended WordNet Knowledge Base. The research addresses the problems of lexical chains ranking and labeling them with appropriate semantic names.

In this paper, we consider the permutation representation of genetic algorithms, and more generally, local search algorithms. We use a variety of permutation distance measures to profile the behavior of the most commonly used mutation operators for permutation-based genetic algorithms. Our operator profiles are also applicable to other local search algorithms, such as simulated annealing, as the most common permutation mutation operators are also commonly found as neighborhood operators for other metaheuristics in a search of the space of permutations. In addition to using several existing distance measures, we introduce two specific instances of the edit distance measure. Our aim is to offer the GA, and local search practitioner, guidance in the selection of mutation and neighborhood operators.