Topo-heuristic reconfiguration of algebraic LSTM components to optimize temporal modulation in long-range dependencies
Petr Andreevich Pylov, Roman Viacheslavovich Maitak, A. V. Dyagileva, Andrey Protodyakonov · E3S Web of Conferences · 2023
In this paper, the authors present a modified LambdaRank ranking algorithm based on the mathematical apparatus of the basic machine learning model (LSTM). LambdaRank is an effective method for ranking objects according to their importance, so it is often used as a mandatory component in search engines and recommendation systems. In this paper, an improvement of the algorithm is proposed by using optimisation techniques and introducing additional parameters for more accurate and stable ranking. The effectiveness of the proposed approach is verified on experimental real application data. The obtained accuracy results of the upgraded algorithm have also been analysed and compared with the classical variation of LambdaRank.