We present a new approach, called First Order Regression (FOR), to handling numerical information in inductive logic programming (ILP). FOR is a combination of ILP and numerical regression. First-order logic descripti...
详细信息
We present a new approach, called First Order Regression (FOR), to handling numerical information in inductive logic programming (ILP). FOR is a combination of ILP and numerical regression. First-order logic descriptions are induced to carve out those subspaces that are amenable to numerical regression among real-valued variables. The program FORS is an implementation of this idea, where numerical regression is focused on a distinguished continuous argument of the target predicate. We show that this can be viewed as a generalisation of the usual ILP problem. Applications of FORS On several real-world data sets are described: the prediction of mutagenicity of chemicals, the modelling of liquid dynamics in a surge tank, predicting the roughness in steel grinding, finite element mesh design, and operator's skill reconstruction in electric discharge machining. A comparison of FORS' performance with previous results in these domains indicates that FORS is an effective tool for ILP applications that involve numerical data.
The automated construction of dynamic system models is an important application area for ILP. We describe a method that learns qualitative models from time-varying physiological signals. The goal is to understand the ...
详细信息
The automated construction of dynamic system models is an important application area for ILP. We describe a method that learns qualitative models from time-varying physiological signals. The goal is to understand the complexity of the learning task when faced with numerical data, what signal processing techniques are required, and how this affects learning. The qualitative representation is based on Kuipers' QSIM. The learning algorithm for model construction is based on Coiera's GENMODEL. We show that QSIM models are efficiently PAC learnable from positive examples only, and that GENMODEL is an ILP algorithm for efficiently constructing a QSIM model. We describe both GENMODEL which performs RLGG on qualitative states to learn a QSIM model, and the front-end processing and segmenting stages that transform a signal into a set of qualitative states. Next we describe results of experiments on data from six cardiac bypass patients. Useful models were obtained, representing both normal and abnormal physiological states. Model variation across time and across different levels of temporal abstraction and fault tolerance is explored. The assumption made by many previous workers that the abstraction of examples from data can be separated from the learning task is not supported by this study. Firstly, the effects of noise in the numerical data manifest themselves in the qualitative examples. Secondly, the models learned are directly dependent on the initial qualitative abstraction chosen.
Interest in research into knowledge discovery in databases (KDD) has been growing continuously because of the rapid increase in the amount of information embedded in real-world data. Several systems have been proposed...
详细信息
Interest in research into knowledge discovery in databases (KDD) has been growing continuously because of the rapid increase in the amount of information embedded in real-world data. Several systems have been proposed for studying the KDD process. One main task in a KDD system is to learn important and user-interesting knowledge from a set of collected data. Most proposed systems use simple machine learning methods to learn the pattern. This may result in efficient performance but the discovery quality is less useful. In this paper, we propose a method to integrated a new and complicated machine learning method called inductive logic programming (ILP) to improve the KDD quality. Such integration shows how this new learning technique can be easily applied to a KDD system and how it can improve the representation of the discovery. In our system, we use user's queries to indicate the importance and interestingness of the target knowledge. The system has been implemented on a SUN workstation using the Sybase database system. Detailed examples are also provided to illustrate the benefit of integrating the ILP technique with the KDD system.
Machine learning can be a most valuable tool for improving the flexibility and efficiency of robot applications. Many approaches to applying machine learning to robotics are known. Some approaches enhance the robot...
详细信息
Machine learning can be a most valuable tool for improving the flexibility and efficiency of robot applications. Many approaches to applying machine learning to robotics are known. Some approaches enhance the robot's high-level processing, the planning capabilities, Other approaches enhance the low-level processing, the control of basic actions. In contrast, the approach presented in this paper uses machine learning for enhancing the link between the low-level representations of sensing and action and the high-level representation of planning. The aim is to facilitate the communication between the robot and the human user. A hierarchy of concepts is learned from route records of a mobile robot. Perception and action are combined at every level, i.e., the concepts are perceptually anchored. The relational learning algorithm GRDT has been developed which completely searches in a hypothesis space, that is restricted by rule schemata, which the user defines in terms of grammars.
Insulin and the insulin-like growth factors IGF-I and IGF-II are found in all vertebrates, and these anabolic peptides share primary and tertiary structural features which suggest that they have evolved from a common ...
详细信息
Insulin and the insulin-like growth factors IGF-I and IGF-II are found in all vertebrates, and these anabolic peptides share primary and tertiary structural features which suggest that they have evolved from a common ancestral gene. We have proposed that an insulin-like peptide (ILP) cDNA recently cloned from the protochodate amphioxus may represent the ancestral gene in that the deduced sequence of ILP contains features of both insulin and IGF, and it evidently represents a hybrid insulin/lGF molecule. To expand this hypothesis we have cloned the cDNA that encodes the cognate receptor from amphioxus. Primary sequence comparisons show that the ILP receptor is a member of the insulin receptor family, which in mammals includes the insulin receptor (IR), type I IGF receptor (IGF-IR), and IR-related receptor (IRR). In overall amino acid sequence, the ILP receptor is 48.6% identical to the human (h)IR, 47.3% identical to hIGF-IR, and 43.7% identical to hIRR, and this contrasts with the finding that hIR and hIGF-IR share 57.6% identity. Using degenerate oligonucleotide primers, we show by RT-PCR that amphioxus contains only a single member of the insulin receptor gene family. To complement the sequence comparison, we expressed the ILP receptor protein by transfecting the cDNA into 293 cells. Autophosphorylation of the expressed ILP receptor was half-maximally stimulated by a synthetic ILP analog, (B1-Thr)ILP, at a concentration of about 5 x 10(-7) M. Interestingly, autophosphorylation of the ILP receptor was also stimulated by incubation with either mammalian insulin or IGF-I, although equally high concentrations (10(-5) M) of each were required. Based on these results, we propose that, analogously to the ILP gene, the ancestral ILP receptor gene also duplicated and diverged to generate the IR and IGF-IR genes during the evolutionary transition from protochordates to vertebrates. Our results also indicate that the amphioxus ILP receptor contains the basic structural deter
We present our machine learning system, that uses inductive logic programming techniques to learn how to identify transmembrane domains from amino acid sequences. Our system facilitates the use of operators such as &#...
详细信息
We present our machine learning system, that uses inductive logic programming techniques to learn how to identify transmembrane domains from amino acid sequences. Our system facilitates the use of operators such as 'contains', that act on entire sequences, rather than on individual elements of a sequence. The prediction accuracy of our new system is around 93%, and this compares favourably with earlier results.
In this article a short review of research and applications in machine learning is given. Rather than attempt to cover all areas of ML, the focus is on its role in building expert systems, its approach to classificati...
详细信息
In this article a short review of research and applications in machine learning is given. Rather than attempt to cover all areas of ML, the focus is on its role in building expert systems, its approach to classification problems and ML methods of learning control. A relatively new area, inductive logic programming, is also discussed.
A common problem in anthropological field work is generalizing rules governing social interactions and relations (particularly kinship) from a series of examples. One class of machine learning algorithms is particular...
详细信息
A common problem in anthropological field work is generalizing rules governing social interactions and relations (particularly kinship) from a series of examples. One class of machine learning algorithms is particularly well-suited to this task: inductive logic programming systems, as exemplified by FOIL. A knowledge base of relationships among individuals is established, in the form of a series of single-predicate facts. Given a set of positive and negative examples of a new relationship, the machine learning programs build a Horn clause description of the target relationship. The power of these algorithms to derive complex hypotheses is demonstrated for a set of kinship relationships drawn from the anthropological literature. FOIL extends the capabilities of earlier anthropology-specific learning programs by providing a more powerful representation for induced relationships, and is better able to learn in the face of noisy or incomplete data.
The ability to learn recursive definitions is a desirable characteristic of a learner. This paper presents Clam, a system that efficiently learns Prolog purely and lefl-recursive definitions from small data sets by us...
详细信息
Knowledge acquisition is a difficult, error-prone, and time-consuming task. The task of automatically improving an existing knowledge base using learning methods is addressed by the class of systems performing theory ...
详细信息
Knowledge acquisition is a difficult, error-prone, and time-consuming task. The task of automatically improving an existing knowledge base using learning methods is addressed by the class of systems performing theory refinement. This paper presents a system, FORTE (First-Order Revision of Theories from Examples), which refines first-order Horn-clause theories by integrating a variety of different revision techniques into a coherent whole. FORTE uses these techniques within a hill-climbing framework, guided by a global heuristic. It identifies possible errors in the theory and calls on a library of operators to develop possible revisions. The best revision is implemented, and the process repeats until no further revisions are possible. Operators are drawn from a variety of sources, including propositional theory refinement, first-order induction, and inverse resolution. FORTE is demonstrated in several domains, including logicprogramming and qualitative modelling.
暂无评论