Federated learning holds great potential for enabling large-scale healthcare research and collaboration across multiple centers while ensuring data privacy and security are not compromised. Although numerous recent studies suggest or utilize federated learning based methods in healthcare, it remains unclear which ones have potential clinical utility. This review paper considers and analyzes the most recent studies up to May 2024 that describe federated learning based methods in healthcare. After a thorough review, we find that the vast majority are not appropriate for clinical use due to their methodological flaws and/or underlying biases which include but are not limited to privacy concerns, generalization issues, and communication costs. As a result, the effectiveness of federated learning in healthcare is significantly compromised. To overcome these challenges, we provide recommendations and promising opportunities that might be implemented to resolve these problems and improve the quality of model development in federated learning with healthcare.

Ruddlesden–Popper (R–P) phase layered chalcogenide perovskites had attracted broad interest as potential lead-free high-performance photovoltaic absorbers. Ca₃Sn₂S₇ is a graphene-like RP phase perovskite with a ultrahigh carrier mobility and a more significant absorption coefficient in the visible light region than those of the classic hybrid halide perovskite MAPbI₃. However, the ultra-low direct bandgap of Ca₃Sn₂S₇ is unfavorable for the photovoltaic application. In this work, we addressed these issues by designing an anion-mixed RP phase perovskite with an appropriate direct bandgap. The idea was to adjust its bandgap with different O proportions from 7.14% to 35.71%. We considered more than 3000 derivative structures of Ca₆Sn₄S_14−xO_x (x = 1–5) that were related to the arrangement of mixed S/O atoms. To ensure that the computational models were based on the screened optimal structures, we found that Ca₆Sn₄S_14−xO_x (x = 4 and 5) could increase the bandgap of Ca₃Sn₂S₇ into the range of 1.19 eV–1.64 eV and 1.02 eV–1.47 eV, respectively. Meanwhile, Ca₆Sn₄S_14−xO_x also had absorption coefficients beyond 10⁵ cm⁻¹. These results made them possible candidates as new-generation photovoltaic absorbers. We also trained the supervised graph convolutional network and the unsupervised Mat-generative adversarial networks (GAN) for accelerating the density functional theory (DFT) calculation of over 3000 structures. Even if considering the time to generate the training samples by DFT, we prove that the Mat-GAN strategy could reduce the DFT calculation consumption by more than 99%. In order to reveal the distributive characteristics of the arrangement of mixed S/O, we adopted active machine learning to analyze the differences of these structures. We found that the O atom would preferentially replace the S in the Sn–S–Sn position.

Background: As large language models continue to expand in size and diversity, their substantial potential and the relevance of their applications are increasingly being acknowledged. The rapid advancement of these models also holds profound implications for the long-term design of stimulus-responsive materials used in drug delivery. Methods: The large model used Hugging Face’s Transformers package with BigBird, Gemma, and GPT NeoX architectures. Pre-training used the PubChem dataset, and fine-tuning used QM7b. Chemist instruction training was based on Direct Preference Optimization. Drug Likeness, Synthetic Accessibility, and PageRank Scores were used to filter molecules. All computational chemistry simulations were performed using ORCA and Time-Dependent Density-Functional Theory. Results: To optimize large models for extensive dataset processing and comprehensive learning akin to a chemist’s intuition, the integration of deeper chemical insights is imperative. Our study initially compared the performance of BigBird, Gemma, GPT NeoX, and others, specifically focusing on the design of photoresponsive drug delivery molecules. We gathered excitation energy data through computational chemistry tools and further investigated light-driven isomerization reactions as a critical mechanism in drug delivery. Additionally, we explored the effectiveness of incorporating human feedback into reinforcement learning to imbue large models with chemical intuition, enhancing their understanding of relationships involving -N=N- groups in the photoisomerization transitions of photoresponsive molecules. Conclusions: We implemented an efficient design process based on structural knowledge and data, driven by large language model technology, to obtain a candidate dataset of specific photoswitchable molecules. However, the lack of specialized domain datasets remains a challenge for maximizing model performance.

Mixed X-anion perovskites, such as CsPbX₃ (X = Cl, Br, or I), play an important role in photovoltaic applications.