Saturday, 16:30 - 18:30 - Room 146

*Stream: Optimization in industry, business and finance*

Chair:

- Optimization Approach for Close to Reality Determination of Stresses in RC Members under Eccentric Compression

A method for determining stresses in the rectangular cross-sections of RC members under eccentric compression is presented. It assumes the nonlinear physical relation for concrete in compression. The task consists of solving the set of nonlinear equations s.t. box constraints. The equations are solved by the least squares method. Modified BFGS quasi-Newton and/or Hooke-Jeeves are applied to find the starting point for the Broyden secant method. This approach is compared with Levenberg-Marquardt method variants. The model is verified on the set of data encountered in engineering practice.

- Randomized methods in solving contemporary traffic engineering problems in telecommunication.

We present several mixed-integer variants of the Multicommodity Flow problem (MCF) coming from the practice of telecommunication. They account for k-splittability of the flows (sending a commodity via at most k paths), lower bound for a path flow (to satisfy a QoS demand for a single connection), reliability (at least 2 disjoint paths for a commodity), multicasting (a possibility of duplicating a flow in a node, modifies the Kirchoff law). We propose and analyze some adaptations of the Raghavan and Thompson's randomized rounding method to approximately and fast solve such variants of MCF.

- Portfolio optimization for a Large Investor under Partial Information and Price Impact

We study a portfolio optimization problem for a large investor where the underlying price process is a diffusion affected by a finite-state Markov chain, representing the state of the market, and the portfolio decision of the investor. We obtain results for different impact and utility function choices. Compared to the case without price impact, we conclude that for logarithmic and power utility choices with linear impact function, the resulting value function dominates in any market regime. We extend our analysis to a partial-information setting in which the Markov chain is not observable.

- Efficient optimization of the reward-risk ratio with polyhedral risk measures

In several problems of portfolio selection the reward-risk ratio criterion is optimized to search for a risky portfolio offering the maximum increase of the mean return, compared to the risk-free investment opportunities. The reward-risk ratio optimization with polyhedral risk meausures can be transformed into LP formulations. The corresponding LP models have typically both the number of constraints the number of variables proportional to the number of scenarios. This decrease dramatically their computational efficiency while dealing with real-life financial problems based on advanced simul

- Evaluating appropriate artificial neural network model for forecasting foreign exchange

Present paper aims at forecasting exchange rate of Dollar, Pound, Euro and Japanese Yen in terms of Indian rupee. The daily data of the respective currencies will considered from 1999 own wards. For forecasting the respective variables, we will apply artificial neural network a nonlinear, non parametric and data driven modelling technique.In the study different combination of neural networks with different combinations of input nodes, one hidden node with altered combinations of hidden nodes and various activation functions will be applied to generate ex ante and ex post forecasts.