SB RAS Conference Information System

Reports summary for author: Tsaregorodtsev, Victor

1. Computational and Informational Technologies for Science, Engineering and Education
   Ust–Kamenogorsk, Kazakhstan, September, 8–12, 2003
   • Tsaregorodtsev V. Relation between Lipschitz value of training set and neural networks properties and speed of training (abstract in russian)
   • Tsaregorodtsev V. Decision accuracy bounds in robust error function based on power metric for neural network prediction tasks (abstract in russian)
   • Tsaregorodtsev V. New approach to building neural software package for real industrial problems (abstract in russian)

2. The International Conference on Computational Mathematics ICCM-2004
   June, 21-25, 2004 Novosibirsk, Academgorodok
   • Tsaregorodtsev V. Towards designing of parallel artificial neural networks software packages: a new approach (abstract in russian)

3. International Conference "Computational and Informational Technologies for Research, Engineering and Education"
   Pavlodar, Kazakhstan, September 20 - 22, 2006
   • Tsaregorodtsev V. Optimal meural network structure finding with coordinate-wise directional descent in structure-effecting operations space (abstract in russian)
   • Tsaregorodtsev V. Additional independent variables constuction based on neural network decision sensitivity to initial input variables (abstract in russian)

4. Computational and Informational Technologies for Research, Engineering and Education CIT-2004
   Alma-Ata, Kazakhstan, October, 6 - 10, 2004
   • Tsaregorodtsev V. General inefficiency of batch gradient descent-based methods for neural network training (abstract in russian)
   • Tsaregorodtsev V. High outputs-to-inputs sensitivity of neural classifier may indicate data outliers (abstract in russian)
   • Tsaregorodtsev V. Learning curves help to optimize properties of every single model while boosting (abstract in russian)

5. Computational and Informational Technologies in Science, Engineering and Education (CTMM-2008)
   Almaty, Kazakhstan, September 10 — September 14, 2008
   • Tsaregorodtsev V. Constructive algorithm of multulayer perceptron synthesis
   • Tsaregorodtsev V. Convolutional neural networks with non-trainable convolutional layers (abstract in russian)
   • Tsaregorodtsev V. Autoassociative neural networks for hidden values discovery problem (abstract in russian)
   • Tsaregorodtsev V., Drobushevskaya O. Senitivity and tolerance of Siberian forest species to climatic parameters (abstract in russian)

Summary: 13 reports at 5 conferences

Comments

|Russian Part| |English Part;|

Go to Home Site