Search for: All records

Abstract One of the more popular optimization methods in current use is the Adam optimizer. This is due, at least in part, to its effectiveness as a training algorithm for deep neural networks, which are associated with many machine learning tasks. In this paper, we introduce time delays into the Adam optimizer. Time delays typically have an adverse effect on dynamical systems, including optimizers, slowing the system’s rate of convergence and potentially causing instabilities. However, our numerical experiments indicate that introducing time-delays into the Adam optimizer can significantly improve its performance, resulting in an often much smaller loss-value. Perhaps more surprising is that this improvement often scales with dimension-the higher the dimension the greater the advantage of using time delays in improving loss-values. Along with describing these results we show that, for the time-delays we consider, the temporal complexity of the delayed Adam optimizer remains the same as the undelayed optimizer and that the algorithm’s spatial complexity scales linearly in the length of the largest time-delay. Last, we extend the theory of intrinsic stability to give a criterion under which the minima, either local or global, associated with the delayed Adam optimizer are stable. 

Structures of many complex biological assemblies are increasingly determined using integrative approaches, in which data from multiple experimental methods are combined. A standalone system, called PDB-Dev, has been developed for archiving integrative structures and making them publicly available. Here, the data standards and software tools that support PDB-Dev are described along with the new and updated components of the PDB-Dev data-collection, processing and archiving infrastructure. Following the FAIR (Findable, Accessible, Interoperable and Reusable) principles, PDB-Dev ensures that the results of integrative structure determinations are freely accessible to everyone. 

Limitations in the applicability, accuracy, and precision of individual structure characterization methods can sometimes be overcome via an integrative modeling approach that relies on information from all available sources, including all available experimental data and prior models. The open-source Integrative Modeling Platform (IMP) is one piece of software that implements all computational aspects of integrative modeling. To maximize the impact of integrative structures, the coordinates should be made publicly available, as is already the case for structures based on X-ray crystallography, NMR spectroscopy, and electron microscopy. Moreover, the associated experimental data and modeling protocols should also be archived, such that the original results can easily be reproduced. Finally, it is essential that the integrative structures are validated as part of their publication and deposition. A number of research groups have already developed software to implement integrative modeling and have generated a number of structures, prompting the formation of an Integrative/Hybrid Methods Task Force. Following the recommendations of this task force, the existing PDBx/mmCIF data representation used for atomic PDB structures has been extended to address the requirements for archiving integrative structural models. This IHM-dictionary adds a flexible model representation, including coarse graining, models in multiple states and/or related by time or other order, and multiple input experimental information sources. A prototype archiving system called PDB-Dev ( https://pdb-dev.wwpdb.org ) has also been created to archive integrative structural models, together with a Python library to facilitate handling of integrative models in PDBx/mmCIF format.