Javier Béjar

Generality-based conceptual clustering with probabilistic concepts

Statistical research in clustering has almost universally focused on data sets described by continuous features and its methods are difficult to apply to tasks involving symbolic features. In addition, these methods are seldom concerned with helping the user in interpreting the results obtained. Machine learning researchers have developed conceptual clustering methods aimed at solving these problems. Following a long term tradition in AI, early conceptual clustering implementations employed logic as the mechanism of concept representation. However, logical representations have been criticized for constraining the resulting cluster structures to be described by necessary and sufficient conditions. An alternative are probabilistic concepts which associate a probability or weight with each property of the concept definition. In this paper, we propose a symbolic hierarchical clustering model that makes use of probabilistic representations and extends the traditional ideas of specificity-generality typically found in machine learning. We propose a parameterized measure that allows users to specify both the number of levels and the degree of generality of each level. By providing some feedback to the user about the balance of the generality of the concepts created at each level and given the intuitive behavior of the user parameter, the system improves user interaction in the clustering process.

Concept Formation in WWTP by Means of Classification Techniques: ACompared Study

Although activated sludge process is a very widely used biologicalprocess in wastewater treatment plants (WWTP), and there areproperly functioning control loops such as that of dissolved oxygen,in practice, this type of plant requires a major time investment onthe part of the operator, involving many manual operations.Treatment plants work well most of the time, as long as there are not unforeseen occurrences. Normal operatingsituations (generally similar to design conditions) can be treatedmathematically by using efficient control algorithms. However, there aresituations in which the control system cannot properlymanage the plant, and in which the process can only be efficiently managedthanks to the operator‘s experience. This is a case in which aknowledge-based system may be useful. One of the difficulties inherent tothe development of a knowledge-based system is to obtain the knowledge base(i.e., knowledge acquisition), specially whendealing with a wide, complicated and ill-structured)field.Among the aims of this work arethose to show how semi-automatic knowledge acquisition tools could helphuman experts to organize their knowledge about their domain and also, tocompare the power of different approaches of knowledge acquisition) to the same database.In this paper are presented the results obtained fromapplying two different classification techniques to the development of knowledge-bases for the management of an activated sludge process.

Integrating Declarative Knowledge in Hierarchical Clustering Tasks

The capability of making use of existing prior knowledge is an important challenge for Knowledge Discovery tasks. As an unsupervised learning task, clustering appears to be one of the tasks that more benefits might obtain from prior knowledge. In this paper, we propose a method for providing declarative prior knowledge to a hierarchical clustering system stressing the interactive component. Preliminary results suggest that declarative knowledge is a powerful bias in order to improve the quality of clustering in domains were the internal biases of the system are inappropriate or there is not enough evidence in data and that it can lead the system to build more comprehensible clusterings.

A Distributed Control System Based on Agent Architecture for Wastewater Treatment

The development and implementation of an agent-based distributed control system in a wastewater treatment plant (WWTP) are shown. The hardware architecture contains different supervision levels, including two autonomous process computers, a programmable logic controller (PLC) and a knowledge-based system (KBS), acting at the top supervisory level. The knowledge is organized in several distributed agents, representing the available knowledge for every subprocess of the WWTP. In addition to these independent agents, a supervisor agent acts as its master. The distributed multi-agent architecture improves previous developments with monolithic KBS, allowing the development of independent and reusable agents. Different strategies have been included, adapted, and modified without a great programming effort. The developed system has involved the transformation of a classical control system into a system that enables a WWTP to respond to different usual problems. Finally, the system has been satisfactorily validated supervising a pilot WWTP for more than 3 years.

Agent Strategies on DPB Auction Tournaments

In this work we present the experience of an electronic tournament between trading agents developed at the Technical University of Catalonia (UPC) as a course work for an artificial intelligence applications course. Using the Fishmarket platform as implementation basis, fourteen different agents were developed and confronted in a set of Downward Bidding Protocol (DBP) auctions in order to measure the performance of their strategies. We present the different groups of strategies of the agents, their architectures and their relationship to the success of the agents.

Reflective Reasoning in a Case-Based Reasoning Agent

As a Case-Based Reasoning agent (CBR) evolves over time, and solves new problems based on previous experiences, there are some pitfalls that can appear in the problem-solving task. When those troubles arise, is the time to start some reflective reasoning tasks to overcome those problems and to improve the CBR performance. Our proposal is to extend the basic reasoning and learning cycle with some new added reflective tasks such as forgetting cases, learning new cases, updating the case library organisation or re-exploring the case library, and including other strategies such as building meta-cases.

Efficient Construction of Comprehensible Hierarchical Clusterings

Clustering is an important data mining task which helps in finding useful patterns to summarize the data. In the KDD context, data mining is often used for description purposes rather than for prediction. However, it turns out difficult to find clustering systems that help to ease the interpretation task to the user in both, statistics and Machine Learning fields. In this paper we present Isaac, a hierarchical clustering system which employs traditional clustering ideas combined with a feature selection mechanism and heuristics in order to provide comprehensible results. At the same time, it allows to efficiently deal with large datasets by means of a preprocessing step. Results suggest that these aims are achieved and encourage further research.

Providing wastewater treatment plants with predictive knowledge based on transition networks

Presents a progress report on integrating predictive skills into an integrated AI system for wastewater treatment plant (WWTP) supervision and control. Although the embedded approaches within the previously developed architecture, called DAI-DEPUR, such as numerical control knowledge, rule-based reasoning and case-based reasoning, are able to cope with the overall supervision task of a plant, one feature is missing: predictive knowledge. With the previous approaches, the supervisory system works reasonably well, but the actuation process always restores the normal operation of a WWTP tardily. Thus, the supervision is implemented in an a posteriori fashion, which can be very dangerous for the environment. The integration of a new kind of knowledge can overcome this problem of control systems.

To Bid or Not To Bid Agent Strategies in Electronic Auction Games

This paper presents the results and analysis of the Fishmarket tournament held this spring at the Technical University of Catalonia (UPC) by a group of undergraduate students as a course work for an artificial intelligence applications course. In the tournament participated sixteen different agents that competed in a three phase eliminatory competition. The agents were divided in groups of four and competed in a number of Downward Bidding Protocol (DBP) auctions for boxes of fish. We present the information analyzed by the students in order to build their agents, what information was considered relevant, and the different strategies of the agents.

A machine learning methodology for the selection and classification of spontaneous spinal cord dorsum potentials allows disclosure of structured (non-random) changes in neuronal connectivity induced by nociceptive stimulation.

Previous studies aimed to disclose the functional organization of the neuronal networks involved in the generation of the spontaneous cord dorsum potentials (CDPs) generated in the lumbosacral spinal segments used predetermined templates to select specific classes of spontaneous CDPs. Since this procedure was time consuming and required continuous supervision, it was limited to the analysis of two specific types of CDPs (negative CDPs and negative positive CDPs), thus excluding potentials that may reflect activation of other neuronal networks of presumed functional relevance. We now present a novel procedure based in machine learning that allows the efficient and unbiased selection of a variety of spontaneous CDPs with different shapes and amplitudes. The reliability and performance of the present method is evaluated by analyzing the effects on the probabilities of generation of different classes of spontaneous CDPs induced by the intradermic injection of small amounts of capsaicin in the anesthetized cat, a procedure known to induce a state of central sensitization leading to allodynia and hyperalgesia. The results obtained with the selection method presently described allowed detection of spontaneous CDPs with specific shapes and amplitudes that are assumed to represent the activation of functionally coupled sets of dorsal horn neurones that acquire different, structured configurations in response to nociceptive stimuli. These changes are considered as responses tending to adequate transmission of sensory information to specific functional requirements as part of homeostatic adjustments.