A biblioteca ibu-boost implementa uma nova abordagem para seleção de splits em árvores GBDT, substituindo a comparação relativa por uma rejeição baseada em limiar absoluto. Isso visa eliminar a necessidade de parâmetros arbitrários como min_gain_to_split, inspirando-se na ideia de "screening transform" de um artigo de pesquisa.
CAWI proposes a new weight initialization framework for Randomized Neural Networks (RdNNs) that addresses the limitation of conventional random initialization ignoring inter-feature dependence. It uses a data-fitted copula to ensure frozen projections respect empirical dependence, improving conditioning and predictive performance.
This paper introduces Regularized Emphatic Temporal-Difference Learning (RETD) to address the stability, projection geometry, and variance trade-off in off-policy temporal-difference learning. It proposes a method that regularizes the auxiliary centering recursion to maintain the positive-definiteness of the ETD key matrix and proves its convergence.
The SEMO algorithm, developed by DeepoMe, addresses the question of identifying measurable biological mismatches for personalized interventions in longevity technology. It acts as a network-medicine algorithmic framework, connecting omics signals and compound target networks, as introduced in a bioRxiv preprint for COVID-19 applications.