TY - JOUR AU - Udvarnoki, Zoltán András AU - Fáth, Gábor AU - Fogarasi, Norbert TI - Quantum advantage of Monte Carlo option pricing JF - JOURNAL OF PHYSICS COMMUNICATIONS J2 - J PHYSICS COMM VL - 7 PY - 2023 IS - 5 SP - 055001 SN - 2399-6528 DO - 10.1088/2399-6528/acd2a4 UR - https://m2.mtmt.hu/api/publication/33831953 ID - 33831953 N1 - Funding Agency and Grant Number: Ministry of Culture and Innovation; National Research, Development and Innovation Office within the Quantum Information National Laboratory of Hungary [2022-2.1.1-NL-2022-00004]; Ministry of Innovation and Technology from the National Research, Development and Innovation Fund of Hungary Funding text: This research was supported by the Ministry of Culture and Innovation and the National Research, Development and Innovation Office within the Quantum Information National Laboratory of Hungary (Grant No. 2022-2.1.1-NL-2022-00004)Prepared with the professional support of the Doctoral Student Scholarship Program of the Co-operative Doctoral Program of the Ministry of Innovation and Technology financed from the National Research, Development and Innovation Fund of Hungary. AB - Quantum computers have the potential to provide quadratic speedup for Monte Carlo methods currently used in various classical applications. In this work, we examine the advantage of quantum computers for financial option pricing with the Monte Carlo method. Systematic and statistical errors are handled in a joint framework, and a relationship to quantum gate error is established. New metrics are introduced for the assessment of quantum advantage based on sample count and optimized error handling. We implement and analyze a Fourier series based approach and demonstrate its benefit over the more traditional rescaling method in function approximation. Our numerical calculations reveal the unpredictable nature of systematic errors, making consistent quantum advantage difficult with current quantum hardware. Our results indicate that very low noise levels, a two-qubit gate error rate below 10 −6 , are necessary for the quantum method to outperform the classical one, but a low number of logical qubits (ca. 20) may be sufficient to see quantum advantage already. LA - English DB - MTMT ER - TY - JOUR AU - Rácz, Attila József AU - Fogarasi, Norbert TI - Improved sparse mean reverting portfolio selection using Simulated Annealing and Extreme Learning Machine JF - ACTA UNIVERSITATIS SAPIENTIAE INFORMATICA J2 - ACTA UNIV SAP INFORM PY - 2022 SN - 1844-6086 UR - https://m2.mtmt.hu/api/publication/32525405 ID - 32525405 LA - English DB - MTMT ER - TY - JOUR AU - Rácz, Attila József AU - Fogarasi, Norbert TI - Trading sparse, mean reverting portfolios using VAR(1) and LSTM prediction JF - ACTA UNIVERSITATIS SAPIENTIAE INFORMATICA J2 - ACTA UNIV SAP INFORM VL - 13 PY - 2021 IS - 2 SP - 288 EP - 302 PG - 15 SN - 1844-6086 DO - 10.2478/ausi-2021-0013 UR - https://m2.mtmt.hu/api/publication/32808022 ID - 32808022 AB - We investigated the predictability of mean reverting portfolios and the VAR(1) model in several aspects. First, we checked the dependency of the accuracy of VAR(1) model on different data types including the original data itself, the return of prices, the natural logarithm of stock and on the log return. Then we compared the accuracy of predictions of mean reverting portfolios coming from VAR(1) with different generative models such as VAR(1) and LSTM for both online and offline data. It was eventually shown that the LSTM predicts much better than the VAR(1) model. The conclusion is that the VAR(1) assumption works well in selecting the mean reverting portfolio, however, LSTM is a better choice for prediction. With the combined model a strategy with positive trading mean profit was successfully developed. We found that online LSTM outperforms all VAR(1) predictions and results in a positive expected profit when used in a simple trading algorithm. LA - English DB - MTMT ER - TY - JOUR AU - Ceffer, Attila AU - Fogarasi, Norbert AU - Levendovszky, János TI - Trading by estimating the quantized forward distribution JF - APPLIED ECONOMICS J2 - APPL ECON VL - 50 PY - 2018 IS - 59 SP - 6397 EP - 6405 PG - 9 SN - 0003-6846 DO - 10.1080/00036846.2018.1486021 UR - https://m2.mtmt.hu/api/publication/3393080 ID - 3393080 LA - English DB - MTMT ER - TY - JOUR AU - Ceffer, Attila AU - Levendovszky, János AU - Fogarasi, Norbert TI - Applying Independent Component Analysis and Predictive Systems for Algorithmic Trading JF - COMPUTATIONAL ECONOMICS J2 - COMPUT ECON VL - 54 PY - 2018 IS - 1 SP - 281 EP - 303 PG - 23 SN - 0927-7099 DO - 10.1007/s10614-017-9719-z UR - https://m2.mtmt.hu/api/publication/3256024 ID - 3256024 N1 - WoS:hiba:000467688100013 2022-12-20 23:00 év nem egyezik LA - English DB - MTMT ER - TY - JOUR AU - Ivanyi, Antal AU - Fogarasi, Norbert TI - On partial sorting in restricted rounds. Dedicated to the memory of Antal Ivanyi TS - Dedicated to the memory of Antal Ivanyi JF - ACTA UNIVERSITATIS SAPIENTIAE INFORMATICA J2 - ACTA UNIV SAP INFORM VL - 9 PY - 2017 IS - 1 SP - 17 EP - 34 PG - 18 SN - 1844-6086 DO - 10.1515/ausi-2017-0002 UR - https://m2.mtmt.hu/api/publication/3251573 ID - 3251573 LA - English DB - MTMT ER - TY - CONF AU - Ceffer, Attila AU - Levendovszky, János AU - Fogarasi, Norbert TI - Applying ICA and NARX networks for algorithmic trading T2 - Fourth International Symposium in Computational Economics and Finance PY - 2016 SP - 1 EP - 16 UR - https://m2.mtmt.hu/api/publication/3183235 ID - 3183235 AB - In this paper, a Nonlinear AutoRegressive network with eXogenous inputs (NARX) is proposed for algorithmic trading by predicting the future value of financial time series. This network is highly capable of modeling vector autoregressive VAR(p) time series. In order to avoid overfitting, the input is pre-processed by Independent Component Analysis (ICA) to filter out the most noise like component. In this way, the accuracy of the prediction is increased. The proposed algorithm has a reduced number of free parameters which makes fast learning and trading possible. The method is not only tested on single asset price series, but also on predicting the value of mean reverting portfolios obtained by maximizing the predictability parameter of VAR(1) processes. The tests were first performed on artificially generated data and then on real data selected from Exchange Traded Fund (ETF) time series including bid-ask spread. In both cases profit has been achieved by the proposed method. LA - English DB - MTMT ER - TY - THES AU - Fogarasi, Norbert TI - Polynomial Time Heuristic Optimization Methods Applied to Problems in Computational Finance PB - Budapesti Műszaki és Gazdaságtudományi Egyetem PY - 2014 SP - 82 UR - https://m2.mtmt.hu/api/publication/2944191 ID - 2944191 LA - English DB - MTMT ER - TY - JOUR AU - Fogarasi, Norbert AU - Levendovszky, János TI - Sparse, mean reverting portfolio selection using simulated annealing JF - ALGORITHMIC FINANCE J2 - ALGORITHM FINANCE VL - 2 PY - 2013 IS - 3-4 SP - 197 EP - 211 PG - 15 SN - 2158-5571 DO - 10.3233/AF-13026 UR - https://m2.mtmt.hu/api/publication/2693507 ID - 2693507 LA - English DB - MTMT ER - TY - JOUR AU - Tornai, Kálmán AU - Fogarasi, Norbert AU - Levendovszky, János TI - Improvements to the Hopfield Neural Network Solution of the Total Weighted Tardiness Scheduling Problem JF - PERIODICA POLYTECHNICA-ELECTRICAL ENGINEERING AND COMPUTER SCIENCE J2 - PERIOD POLYTECH ELECTR ENG COMP SCI VL - 57 PY - 2013 IS - 2 SP - 57 EP - 64 PG - 8 SN - 2064-5260 DO - 10.3311/PPee.2090 UR - https://m2.mtmt.hu/api/publication/2387414 ID - 2387414 AB - This paper explores novel, polynomial time, heuristic, approximate solutions to the NP-hard problem of finding the optimal job schedule on identical machines which minimizes total weighted tardiness (TWT). We map the TWT problem to quadratic optimization and demonstrate that the Hopfield Neural Network (HNN) can successfully solve it. Furthermore, the solution can be significantly sped up by choosing the initial state of the HNN as the result of a known simple heuristic, we call this Smart Hopfield Neural Network (SHNN). We also demonstrate, through extensive simulations, that by considering random perturbations to the Largest Weighted Process First (LWPF) and SHNN methods, we can introduce further improvements to the quality of the solution, we call the latter Perturbed Smart Hopfield Neural Network (PSHNN). Finally, we argue that due to parallelization techniques, such as the use of GPGPU, the additional cost of these improvements is small. Numerical simulations demonstrate that PSHNN outperforms HNN in over 99% of all randomly generated cases by an average of 3-7%, depending on the problem size. On a specific, large scale scheduling problem arising in computational finance at Morgan Stanley, one of the largest financial institutions in the world, PSHNN produced a 5% improvement over the next best heuristic. LA - English DB - MTMT ER -