Muggleton, S., A. Srinivasan, R. D. King and M. J. E. Sternberg #1998#. Biochemical knowledge discovery using inductive logic programming. In: Proceedings of the 1st International Conference on Discovery Science #DS-98# #S. Arikawa and H. Motoda, Eds.#. Vol. 1532 of LNAI. Springer. Berlin. pp. 326#341.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....of the learned concepts. The use of programs from the logic of rst order as underlying representation makes ILP systems more ecient and more useful than conventional empirical machine learning systems. ILP systems were used successfully for the solution of several problems of the real world ( 10] [11]) Most of the representations in rst order logic, which are used in ILP systems, are versions of the Horn clauses based on Prolog. The two most used operators (see Figure 2) with which consistent programs can be generated are specialization and generalization. Consistent is the fact that a ....

S. Muggleton, A. Srinivasan, R. King, and M. Sternberg. Biochemical knowledge discovery using Inductive Logic Programming. In H. Motoda, editor, Proc. of the rst Conference on Discovery Science, Berlin, 1998. Springer-Verlag.

....lower on the first three targets than the best reachable accuracies; however, 1BC achieves a considerably better result than the best rule inducer predicting the good reversal of scopolamine induced memory deficiency. 6.3. Mutagenesis This problem concerns identifying mutagenic compounds [21, 17]. We considered the regression friendly dataset. In these experiments, we used the atom and bond structure of the molecule as one setting, adding the lumo and logp properties to get a second setting, FIRST ORDER BAYESIAN CLASSIFICATION WITH 1BC 23 and finally adding boolean indicators I a and I ....

....features used at level 1 decomposition is: mol2atom(M,A) atomeq(A,c22 0.117) The accuracies are reported on the first line (level 1) of the Table 3. All accuracies are evaluated using a 10 fold cross validation. Table 3. Results in the mutagenesis domain. The last three lines are taken from [17]. Regression did not use the atoms and bonds knowledge. System Atoms and bonds only Plus lumo and logp Plus I 1 and I a Level 1 81.4 82.4 85.1 Level 2 80.3 82.4 87.2 Progol 88 88 Regression 85 89 Default 66 66 Level 2 decomposition consists in splitting the 3 tuple representing ....

S. Muggleton, A. Srinivasan, R. King, and M. Sternberg. Biochemical knowledge discovery using Inductive Logic Programming. In H. Motoda, editor, Proceedings of the first Conference on Discovery Science, Berlin, 1998. Springer-Verlag.

....p(cjcond) Each of them can be also used for calculating the same term in WRAcc. 3 Experiments We performed experiments on a collection of twenty one domains from the UCI Repository of Machine Learning Databases and Domain Theories [5] and two datasets originating from mutagenesis domain [4]. These domains have been widely used in other comparative studies. The domains properties (number of classes, number of examples, number of discrete, continuous and all attributes and class distribution) are given in Table 1. Table 1: Properties of the domains Dataset nc ne nda nca Class ....

Muggleton, S., Srinivasan, A., King R. and Sternberg, M. (1998) Biochemical knowledge discovery using Inductive Logic Programming. In Motoda, H. (editor) Proceedings of the first Conference on Discovery Science. Springer-Verlag.

....This motivates a reappraisal of flattened Datalog representations which remain closer to the Entity Relationship model. The learning tasks we consider are knowledge discovery and classification on structured data. Data we consider are for instance molecules. In the mutagenicity problem [SMKS94,MSKS98] the example consists of molecules, the target is to predict whether a molecule is mutagenic. A molecule is described by a set of results to some chemical tests, and also by its set of atoms and bonds. There is no straightforward propositional representation of molecules, and thus the problem is ....

S. Muggleton, A. Srinivasan, R. King, and M. Sternberg. Biochemical knowledge discovery using Inductive Logic Programming. In H. Motoda, editor, Proceedings of the first Conference on Discovery Science, Berlin, 1998. Springer-Verlag.

....bias, the following simple rules are induced: eastbound(T,true) hasCar(T,C) clength(C,short) not croof(C,no) eastbound(T,false) not (hasCar(T,C) clength(C,short) not croof(C,no) That is, a train is eastbound if and only if it has a short closed car. The mutagenesis learning task [42] concerns predicting which molecular compounds cause DNA mutations. The mutagenesis dataset consists of 230 classified molecules; 188 of these have been found to be amenable to regression modelling, and the remaining 42, to which we restrict attention here, as regressionunfriendly . The dataset ....

....regression modelling, and the remaining 42, to which we restrict attention here, as regressionunfriendly . The dataset furthermore includes two hand crafted indicator attributes I 1 and I a to introduce some degree of structural detail into the regression equation; following some experiments in [42] we did not include these indicators. Experiment 2 (LINUS applied to mutagenesis) We ran LINUS on the 42 regressionunfriendly molecules, using a non determinate background theory consisting of all 57 first order features with one utility literal concerning atoms (i.e. discarding bond ....

[Article contains additional citation context not shown here]

S. Muggleton, A. Srinivasan, R. King, and M. Sternberg. Biochemical knowledge discovery using Inductive Logic Programming. In H. Motoda, editor, Proceedings of the first Conference on Discovery Science, Berlin, 1998. Springer-Verlag. 40

....the number of learned rules, at the expense of (on average) a small drop in accuracy. 4 Experiments We performed experiments on a collection of twenty one domains from the UCI Repository of Machine Learning Databases and Domain Theories [6] and two data sets originating from mutagenesis domain [5]. These domains have been widely used in other comparative studies. The domains properties (number of examples, number of discrete and continuous attributes and class distribution) are given in Table 1. Performance of the rule inducing algorithms were measured using 10 fold stratified cross ....

Muggleton, S., Srinivasan, A., King R. and Sternberg, M. (1998) Biochemical knowledge discovery using Inductive Logic Programming. In Motoda, H. (editor) Proceedings of the first Conference on Discovery Science. Springer-Verlag.

....categories: Given events and their duration together with a classification in terms of a category c of the next higher level, learn the definition of c. Non determinate sequence prediction has been solved by [25] and has currently received attention in the context of biochemical analyses [31]. It is also the task that has to be solved for language learning. Since the datasets for sequence prediction do not include any explicit time stamp, we do exclude this very interesting issue here. A more detailed structure of time phenomena distinguishes between handling abstract and actual ....

S. Muggleton, A. Srinivasan, R. King, and M. Sternberg. Biochemical knowledge discovery using inductive logic programming. In Hiroshi Motoda, editor, Procs. First International Conference on Discovery Science. Springer, 1998.

....allowed within the feature bias, the following simple rules are induced: east(T) hasCar(T,C) clength(C,short) not croof(C,no) west(T) not (hasCar(T,C) clength(C,short) not croof(C,no) That is, a train is eastbound if and only if it has a short closed car. The mutagenesis learning task [15] concerns predicting which molecular compounds cause DNA mutations. The mutagenesis dataset consists of 230 classified molecules; 188 of these have been found to be amenable to regression modelling, and the remaining 42, to which we restrict attention here, as regression unfriendly . The dataset ....

....regression modelling, and the remaining 42, to which we restrict attention here, as regression unfriendly . The dataset furthermore includes two hand crafted indicator attributes I 1 and I a to introduce some degree of structural detail into the regression equation; following some experiments in [15] we did not include these indicators. Experiment 2 (LINUS applied to mutagenesis) We ran LINUS on the 42 regressionunfriendly molecules, using a non determinate background theory consisting of all 57 first order features with one utility literal concerning atoms (i.e. discarding bond ....

S. Muggleton, A. Srinivasan, R. King, and M. Sternberg. Biochemical knowledge discovery using Inductive Logic Programming. In H. Motoda, editor, Proceedings of the first Conference on Discovery Science, Berlin, 1998. Springer-Verlag.

No context found.

S.H. Muggleton, A. Srinivasan, R.D. King, and M.J.E. Sternberg. Biochemical knowledge discovery using Inductive Logic Programming. In H. Motoda, editor, First Conference on Discovery Science. Springer-Verlag, 1998.

.... between researchers now variously located at the Universities of Edinburgh, Louisville, Oxford, Wales, and York; the Imperial Cancer Research Fund (ICRF) Pfizer UK; and Smith Kline Beecham has resulted in applications of symbolic machine learning to problems in molecular biology and biochemistry [6 14, 23, 26, 31, 33]. Much of this has been accomplished within the setting of Inductive Logic Programming (ILP: see [21] This is an anecdotal account of some practical guidelines that I have found useful during the course of the applied work. That they have had a role to play in my thinking about the biological ....

S.H. Muggleton, A. Srinivasan, R.D. King, and M.J.E. Sternberg. Biochemical knowledge discovery using Inductive Logic Programming. In H. Motoda, editor, First Conference on Discovery Science. Springer-Verlag, 1998.

....given by the pair of numbers (A; T ) denoting respectfully, estimates of the predictive accuracy of the model constructed and the time taken to construct the model 3 . 3 Explanatory value is an important attribute for any model, and should warrant inclusion in the performance measure. In [18] the explanatory value of a model is newpaper.tex; 6 07 2001; 20:11; p.6 generalise 00 (B; P; I ; L; E) Given background knowledge B, a partialordering P over equivalence classes of predicates in B, hypothesis constraints I , a nite training set E = E [ E , returns a model H in L such ....

S.H. Muggleton, A. Srinivasan, R.D. King, and M.J.E. Sternberg. Biochemical knowledge discovery using Inductive Logic Programming. In H. Motoda, editor, First Conference on Discovery Science. Springer-Verlag, 1998.

No context found.

Muggleton, S., A. Srinivasan, R. D. King and M. J. E. Sternberg #1998#. Biochemical knowledge discovery using inductive logic programming. In: Proceedings of the 1st International Conference on Discovery Science #DS-98# #S. Arikawa and H. Motoda, Eds.#. Vol. 1532 of LNAI. Springer. Berlin. pp. 326#341.

No context found.

Muggleton, S., Srinivasan, A., King R. and Sternberg, M. #1998# Biochemical knowledge discovery using Inductive Logic Programming. In Motoda, H. #editor# Proceedings of the #rst Conference on Discovery Science. Springer-Verlag.

No context found.

Muggleton, S., Srinivasan, A., King, R., & Sternberg, M. (1998). Biochemical knowledge discovery using Inductive Logic Programming. Proceedings of the first Conference on Discovery Science. Berlin: Springer-Verlag.

No context found.

Muggleton S., Srinivasan A., King R. & Sternberg M.: Biochemical knowledge discovery using Inductive Logic Programming. In Proceedings of the first Conference on Discovery Science, Motoda H. (ed), Springer-Verlag, Berlin, 1998.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC