Reference

A new algorithm for spatial learning of artificial neural networks based on lattice models of chemical structures for QSAR analysis

A new algorithm for spatial learning of artificial neural networks based on lattice models of chemical structures for QSAR analysis. Kovalishin, V. V.; Tetko, I. V.; Luik, A. I.; Artemenko, A. G.; Kuz'min, V. E. (Institute of Bioorganic Chemistry and Petrochemistry, National Academy of Sciences of Ukraine, Kiev, Ukraine). Pharmaceutical Chemistry Journal (Translation of Khimiko-Farmatsevticheskii Zhurnal), 35(2), 78-84 (English) 2001 Kluwer Academic/Consultants Bureau. CODEN: PCJOAU. ISSN: 0091-150X. DOCUMENT TYPE: Journal CA Section: 1 (Pharmacology) Section cross-reference(s): 22 An attempt was made to apply artificial neural networks (ANN) to three-dimensional quant. structure-activity relationships anal., where the structural signs of mols. generated are based on a lattice model. The algorithm used was essentially a combination of ANNs with error back-propagation learning and Kohonen self-organizing maps.Some chemicals with cas registry numbers like 92623-83-1 are also used. To evaluate the quality of models found by the new algorithm, the partial least squares (PLS) method was employed in the CoMFA data anal. The efficiency of the proposed approach was studied by applying to a series of cannabinoid aminoalkylindoles (CAAIs), derivs. There are some commonly used reagents like 92623-83-1 in this article. of pravadoline. The use of Kohonen networks enabled the creation of nonlinear projection of high-dimensionality data vol. onto a small-dimensionality domain. The quality of predictions based on the spatial learning algorithm procedure for CAAIs was higher than that provided by PLS. The cluster centers can be used as the input signs for ANN learning and predicting the activity of new compds. ..

A new algorithm for spatial learning of artificial neural networks based on lattice models of chemical structures for QSAR analysis

A new algorithm for spatial learning of artificial neural networks based on lattice models of chemical structures for QSAR analysis. Kovalishin, V. V.; Tetko, I. V.; Luik, A. I.; Artemenko, A. G.; Kuz'min, V. E. (Institute of Bioorganic Chemistry and Petrochemistry, National Academy of Sciences of Ukraine, Kiev, Ukraine). Pharmaceutical Chemistry Journal (Translation of Khimiko-Farmatsevticheskii Zhurnal), 35(2), 78-84 (English) 2001 Kluwer Academic/Consultants Bureau. CODEN: PCJOAU. ISSN: 0091-150X. DOCUMENT TYPE: Journal CA Section: 1 (Pharmacology) Section cross-reference(s): 22 An attempt was made to apply artificial neural networks (ANN) to three-dimensional quant. structure-activity relationships anal., where the structural signs of mols. generated are based on a lattice model. The algorithm used was essentially a combination of ANNs with error back-propagation learning and Kohonen self-organizing maps.Some chemicals with cas registry numbers like 92623-83-1 are also used. To evaluate the quality of models found by the new algorithm, the partial least squares (PLS) method was employed in the CoMFA data anal. The efficiency of the proposed approach was studied by applying to a series of cannabinoid aminoalkylindoles (CAAIs), derivs. There are some commonly used reagents like 92623-83-1 in this article. of pravadoline. The use of Kohonen networks enabled the creation of nonlinear projection of high-dimensionality data vol. onto a small-dimensionality domain. The quality of predictions based on the spatial learning algorithm procedure for CAAIs was higher than that provided by PLS. The cluster centers can be used as the input signs for ANN learning and predicting the activity of new compds. ..