Search for: All records

Deploying complex machine learning models on resource-constrained devices is challenging due to limited computational power, memory, and model retrainability. To address these limitations, a hybrid system can be established by augmenting the local model with a server-side model, where samples are selectively deferred by a rejector and then sent to the server for processing. The hybrid system enables efficient use of computational resources while minimizing the overhead associated with server usage. The recently proposed Learning to Help (L2H) model proposed training a server model given a fixed local (client) model. This differs from the Learning to Defer (L2D) framework which trains the client for a fixed (expert) server. In both L2D and L2H, the training includes learning a rejector at the client to determine when to query the server. In this work, we extend the L2H model from binary to multi-class classification problems and demonstrate its applicability in a number of different scenarios of practical interest in which access to the server may be limited by cost, availability, or policy. We derive a stage-switching surrogate loss function that is differentiable, convex, and consistent with the Bayes rule corresponding to the 0-1 loss for the L2H model. Experiments show that our proposed methods offer an efficient and practical solution for multi-class classification in resource-constrained environments. 

Abstract Open access to three-dimensional atomic-level biostructure information from the Protein Data Bank (PDB) facilitated discovery/development of 100% of the 34 new low molecular weight, protein-targeted, antineoplastic agents approved by the US FDA 2019–2023. Analyses of PDB holdings, the scientific literature, and related documents for each drug-target combination revealed that the impact of structural biologists and public-domain 3D biostructure data was broad and substantial, ranging from understanding target biology (100% of all drug targets), to identifying a given target as likely druggable (100% of all targets), to structure-guided drug discovery (>80% of all new small-molecule drugs, made up of 50% confirmed and >30% probable cases). In addition to aggregate impact assessments, illustrative case studies are presented for six first-in-class small-molecule anti-cancer drugs, including a selective inhibitor of nuclear export targeting Exportin 1 (selinexor, Xpovio), an ATP-competitive CSF-1R receptor tyrosine kinase inhibitor (pexidartinib,Turalia), a non-ATP-competitive inhibitor of the BCR-Abl fusion protein targeting the myristoyl binding pocket within the kinase catalytic domain of Abl (asciminib, Scemblix), a covalently-acting G12C KRAS inhibitor (sotorasib, Lumakras or Lumykras), an EZH2 methyltransferase inhibitor (tazemostat, Tazverik), and an agent targeting the basic-Helix-Loop-Helix transcription factor HIF-2α (belzutifan, Welireg). 

A fault-tolerant quantum computer must decode and correct errors faster than they appear. The faster errors can be corrected, the more time the computer can do useful work. The Union-Find (UF) decoder is promising with an average time complexity slightly higher than O(d3). We report a distributed version of the UF decoder that exploits parallel computing resources for further speedup. Using an FPGA-based implementation, we empirically show that this distributed UF decoder has a sublinear average time complexity with regard to d, given O(d3) parallel computing resources. The decoding time per measurement round decreases as d increases, a first time for a quantum error decoder. The implementation employs a scalable architecture called Helios that organizes parallel computing resources into a hybrid tree-grid structure. We are able to implement d up to 21 with a Xilinx VCU129 FPGA, for which an average decoding time is 11.5 ns per measurement round under phenomenological noise of 0.1%, significantly faster than any existing decoder implementation. Since the decoding time per measurement round of Helios decreases with d, Helios can decode a surface code of arbitrarily large d without a growing backlog. 

Abstract The Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB, RCSB.org), the US Worldwide Protein Data Bank (wwPDB, wwPDB.org) data center for the global PDB archive, provides access to the PDB data via its RCSB.org research-focused web portal. We report substantial additions to the tools and visualization features available at RCSB.org, which now delivers more than 227000 experimentally determined atomic-level three-dimensional (3D) biostructures stored in the global PDB archive alongside more than 1 million Computed Structure Models (CSMs) of proteins (including models for human, model organisms, select human pathogens, crop plants and organisms important for addressing climate change). In addition to providing support for 3D structure motif searches with user-provided coordinates, new features highlighted herein include query results organized by redundancy-reduced Groups and summary pages that facilitate exploration of groups of similar proteins. Newly released programmatic tools are also described, as are enhanced training opportunities.