Search for: All records

Deep neural network models provide a powerful experimental platform for exploring core mechanisms underlying human visual perception, such as perceptual grouping and contour integration—the process of linking local edge elements to arrive at a unified perceptual representation of a complete contour. Here, we demonstrate that feedforward convolutional neural networks (CNNs) fine-tuned on contour detection show this human-like capacity, but without relying on mechanisms proposed in prior work, such as lateral connections, recurrence, or top-down feedback. We identified two key properties needed for ImageNet pre-trained, feed-forward models to yield human-like contour integration: first, progressively increasing receptive field structure served as a critical architectural motif to support this capacity; and second, biased fine-tuning for contour-detection specifically for gradual curves (~20 degrees) resulted in human-like sensitivity to curvature. We further demonstrate that fine-tuning ImageNet pretrained models uncovers other hidden human-like capacities in feed-forward networks, including uncrowding (reduced interference from distractors as the number of distractors increases), which is considered a signature of human perceptual grouping. Thus, taken together these results provide a computational existence proof that purely feedforward hierarchical computations are capable of implementing gestalt “good continuation” and perceptual organization needed for human-like contour-integration and uncrowding. More broadly, these results raise the possibility that in human vision, later stages of processing play a more prominent role in perceptual-organization than implied by theories focused on recurrence and early lateral connections. 

We study the electron addition spectrum of the one-dimensional extended Hubbard-Su-Schrieffer-Heeger (HSSH) model in the dilute limit using the density matrix renormalization group method. In addition to the expected renormalization to the band structure, we find that the electron-phonon (𝑒-ph) interaction produces an anomalous spectral feature when electrons are added in the singlet channel but which is absent in the triplet channel. By comparing these results with those obtained from perturbation theory in the antiadiabatic limit, we demonstrate that this anomalous feature is a remnant of the strong electron-electron interaction mediated by the SSH coupling previously derived in the two-particle limit. By studying the evolution of this feature as a function of doping, we track the fate of this attraction to higher carrier concentrations and provide predictions for the spectral features to help guide future searches for strong 𝑒-ph mediated pairing. 

ABSTRACT Stable isotope probing (SIP) experiments in conjunction with Raman microspectroscopy (Raman) or nano-scale secondary ion mass spectrometry (NanoSIMS) are frequently used to explore single cell metabolic activity in pure cultures as well as complex microbiomes. Despite the increasing popularity of these techniques, the comparability of isotope incorporation measurements using both Raman and NanoSIMS directly on the same cell remains largely unexplored. This knowledge gap creates uncertainty about the consistency of single-cell SIP data obtained independently from each method. Here, we conducted a comparative analysis of 543Escherichia colicells grown in M9 minimal medium in the absence or presence of heavy water (²H₂O) using correlative Raman and NanoSIMS measurements to quantify the results between the two approaches. We demonstrate that Raman and NanoSIMS yield highly comparable measurements of²H incorporation, with varying degrees of similarity based on the mass ratios analyzed using NanoSIMS. The¹²C²H/¹²C¹H and¹²C₂²H/¹²C₂¹H mass ratios provide targeted measurements of C-H bonds but may suffer from biases and background interference, while the²H/¹H ratio captures all hydrogen with lower detection limits, making it suitable for applications requiring comprehensive²H quantification. Importantly, despite its higher mass resolution requirements, the use of C₂²H/C₂¹H may be a viable alternative to the use of C²H/C¹H due to lower background and higher overall count rates. Furthermore, using an empirical approach in determining Raman wavenumber ranges via the second derivative improved the data equivalency of²H quantification between Raman and NanoSIMS, highlighting its potential for enhancing cross-technique comparability. These findings provide a robust framework for leveraging both techniques, enabling informed experimental design and data interpretation. By enhancing cross-technique comparability, this work advances SIP methodologies for investigating microbial metabolism and interactions in diverse systems.IMPORTANCEAccurate and reliable measurements of cellular properties are fundamental to understand the function and activity of microbes. This study addresses to what extent Raman microspectroscopy and nano-scale secondary ion mass spectrometry (NanoSIMS) measurements of single cell anabolic activity can be compared. Here, we study the relationship of the incorporation of a stable isotope (²H through incorporation of²H₂O) as determined by the two techniques and calculate a correlation coefficient to support the use of either technique when analyzing cells incubated with²H₂O. The ability to discern between the comparative strengths and limitations of these techniques is invaluable in refining experimental protocols, enhancing data comparability between studies, data interpretation, and ultimately advancing the quality and reliability of outcomes in microbiome research. 

We report two analytical quantum mechanics (QM) models for approximating appropriately scaled harmonic zero-point energies (ZPEs) without Hessian calculations. Following our earlier bond energies from bond orders and populations model that takes a similar form as an extended Hückel model but uses well-conditioned orbital populations, this work demonstrates a proof of concept for approximating ZPEs, an important component in thermochemistry calculations, while eschewing unfavorably scaling algorithms involving Hessian matrices. The ZPE-BOP1 model uses Mulliken orbital populations from hybrid Kohn–Sham density functional theory calculations within an extended Hückel-type model that defines vibrational bond energy terms using two atom-pairwise parameters that are fit to reproduce ZPEs from B3LYP calculations. The more accurate ZPE-BOP2 model uses Mulliken orbital populations from Hartree–Fock calculations within a different extended Hückel-type model that includes a short-range anharmonic energy term and a coupled three-body oscillator energy term with seven atom-pairwise parameters. Both models predict ZPEs in molecules involving first row elements, but ZPE-BOP2 outperforms ZPE-BOP1 in strained and long-chain molecules and provides ZPEs more competitive with those from semi-empirical QM methods (e.g., AM1, PM6, PM7, and XTB-2) that compute ZPEs with Hessian calculations. This work shows progress and an outlook toward computational models that use well-conditioned orbital populations to efficiently predict useful physicochemical properties. It also shows opportunities for approximate QM models that would shift traditional computational bottlenecks away from costly algorithms such as Hessian calculations to others that focus on reliable orbital populations. 

The ability to recognize hidden symmetry in a highly asymmetric world is a key factor in how we view and understand the world around us. Despite the fact that it is an intrinsic property of the natural world, we have an innate ability to find hidden symmetry in asymmetric objects. The inherent asymmetry of the natural world is a fundamental property built into its chemical building blocks (e.g., proteins, carbohydrates, etc.). This review highlights the role of asymmetry in the structure of the carbohydrates and how these stereochemical complexities present synthetic challenges. This survey starts with an overview of the role synthetic chemistry plays in the discovery of carbohydrates and their 3D structure. This review then introduces various de novo asymmetric synthetic approaches that have been developed for the synthesis of carbohydrates and, in particular, oligosaccharides. The two most successful strategies for oligosaccharide synthesis rely on diastereoselective palladium-catalyzed glycosylation. The first uses an Achmatowicz reaction to asymmetrically prepare pyranose building blocks along with a substrate-controlled Pd-glycosylation. The other strategy couples a ligand-controlled Pd-glycosylation with a ring-closing metathesis for oligosaccharide assembly.