Search for: All records

Abstract Visualizing atomic-orbital degrees of freedom is a frontier challenge in scanned microscopy. Some types of orbital order are virtually imperceptible to normal scattering techniques because they do not reduce the overall crystal lattice symmetry. A good example is d xz / d yz (π,π) orbital order in tetragonal lattices. For enhanced detectability, here we consider the quasiparticle scattering interference (QPI) signature of such (π,π) orbital order in both normal and superconducting phases. The theory reveals that sublattice-specific QPI signatures generated by the orbital order should emerge strongly in the superconducting phase. Sublattice-resolved QPI visualization in superconducting CeCoIn 5 then reveals two orthogonal QPI patterns at lattice-substitutional impurity atoms. We analyze the energy dependence of these two orthogonal QPI patterns and find the intensity peaked near E  = 0, as predicted when such (π,π) orbital order is intertwined with d -wave superconductivity. Sublattice-resolved superconductive QPI techniques thus represent a new approach for study of hidden orbital order. 

Episodic memories are records of personally experienced events, coded neurally via the hippocampus and sur- rounding medial temporal lobe cortex. Information about the neural signal corresponding to a memory representation can be measured in fMRI data when the pattern across voxels is examined. Prior studies have found that similarity in the voxel patterns across repetition of a to-be-remembered stimulus predicts later memory retrieval, but the results are inconsistent across studies. The current study investigates the possibility that cognitive goals (defined here via the task instructions given to participants) during encoding affect the voxel pattern that will later support memory retrieval, and therefore that neural representations cannot be interpreted based on the stimulus alone. The behavioral results showed that exposure to variable cognitive tasks across repetition of events benefited subsequent memory retrieval. Voxel patterns in the hippocampus indicated a significant interaction between cognitive tasks (variable vs. consistent) and memory (remembered vs. forgotten) such that reduced voxel pattern similarity for repeated events with variable cognitive tasks, but not consistent cognitive tasks, sup- ported later memory success. There was no significant interaction in neural pattern similarity between cognitive tasks and memory success in medial temporal cortices or lateral occipital cortex. Instead, higher similarity in voxel patterns in right medial temporal cortices was associated with later memory retrieval, regardless of cognitive task. In conclusion, we found that the relationship between pattern similarity across repeated encoding and memory success in the hippocampus (but not medial temporal lobe cortex) changes when the cognitive task during encoding does or does not vary across repetitions of the event. 

Scholarship focused on Black male students in school counseling has been intermittent despite being well documented in the larger field of education and other disciplines. In this article, we conducted a systematic review of the school counseling literature that focused on Black male students. We used critical race theory (CRT) to examine the programs and interventions that have been published with Black male participants in school settings within the school counseling literature and examined the role that school counselors took when supporting Black male students’ academic, social emotional, college and career identity development. We re-conceptualize the Achieving Success Everyday (ASE) group model (Steen et al., 2014) and call for others to use the ASE group model to combat racism and foster Black excellence. 

Deep neural networks (DNNs) achieve state-of-theart performance in many areas, including computer vision, system configuration, and question-answering. However, DNNs are expensive to develop, both in intellectual effort (e.g., devising new architectures) and computational costs (e.g., training). Reusing DNNs is a promising direction to amortize costs within a company and across the computing industry. As with any new technology, however, there are many challenges in re-using DNNs. These challenges include both missing technical capabilities and missing engineering practices. This vision paper describes challenges in current approaches to DNN re-use. We summarize studies of re-use failures across the spectrum of re-use techniques, including conceptual (e.g., reusing based on a research paper), adaptation (e.g., re-using by building on an existing implementation), and deployment (e.g., direct re-use on a new device). We outline possible advances that would improve each kind of re-use. 

Mobile and embedded devices are becoming ubiquitous. Applications such as rescue with autonomous robots and event analysis on traffic cameras rely on devices with limited power supply and computational sources. Thus, the demand for efficient computer vision algorithms increases. Since 2015, we have organized the IEEE Low-Power Computer Vision Challenge to advance the state of the art in low-power computer vision. We describe the competition organizing details including the challenge design, the reference solution, the dataset, the referee system, and the evolution of the solutions from two winning teams. We examine the winning teams’ development patterns and design decisions, focusing on their techniques to balance power consumption and accuracy. We conclude that a successful competition needs a well-designed reference solution and automated referee system, and a solution with modularized components is more likely to win. We hope this paper provides guidelines for future organizers and contestants of computer vision competitions. 

Regular expressions are used for diverse purposes, including input validation and firewalls. Unfortunately, they can also lead to a security vulnerability called ReDoS (Regular Expression Denial of Service), caused by a super-linear worst-case execution time during regex matching. Due to the severity and prevalence of ReDoS, past work proposed automatic tools to detect and fix regexes. Although these tools were evaluated in automatic experiments, their usability has not yet been studied; usability has not been a focus of prior work. Our insight is that the usability of existing tools to detect and fix regexes will improve if we complement them with anti-patterns and fix strategies of vulnerable regexes. We developed novel anti-patterns for vulnerable regexes, and a collection of fix strategies to fix them. We derived our anti-patterns and fix strategies from a novel theory of regex infinite ambiguity — a necessary condition for regexes vulnerable to ReDoS. We proved the soundness and completeness of our theory. We evaluated the effectiveness of our anti-patterns, both in an automatic experiment and when applied manually. Then, we evaluated how much our anti-patterns and fix strategies improve developers’ understanding of the outcome of detection and fixing tools. Our evaluation found that our anti-patterns were effective over a large dataset of regexes (N=209,188): 100% precision and 99% recall, improving the state of the art 50% precision and 87% recall. Our anti-patterns were also more effective than the state of the art when applied manually (N=20): 100% developers applied them effectively vs. 50% for the state of the art. Finally, our anti-patterns and fix strategies increased developers’ understanding using automatic tools (N=9): from median “Very weakly” to median “Strongly” when detecting vulnerabilities, and from median “Very weakly” to median “Very strongly” when fixing them. 

Due to the cost of developing and training deep learning models from scratch, machine learning engineers have begun to reuse pre-trained models (PTMs) and fine-tune them for downstream tasks. PTM registries known as "model hubs" support engineers in distributing and reusing deep learning models. PTM packages include pre-trained weights, documentation, model architectures, datasets, and metadata. Mining the information in PTM packages will enable the discovery of engineering phenomena and tools to support software engineers. However, accessing this information is difficult - there are many PTM registries, and both the registries and the individual packages may have rate limiting for accessing the data. We present an open-source dataset, PTMTorrent, to facilitate the evaluation and understanding of PTM packages. This paper describes the creation, structure, usage, and limitations of the dataset. The dataset includes a snapshot of 5 model hubs and a total of 15,913 PTM packages. These packages are represented in a uniform data schema for cross-hub mining. We describe prior uses of this data and suggest research opportunities for mining using our dataset. 

Deep Neural Networks (DNNs) are being adopted as components in software systems. Creating and specializing DNNs from scratch has grown increasingly difficult as stateof- the-art architectures grow more complex. Following the path of traditional software engineering, machine learning engineers have begun to reuse large-scale pre-trained models (PTMs) and fine-tune these models for downstream tasks. Prior works have studied reuse practices for traditional software packages to guide software engineers towards better package maintenance and dependency management. We lack a similar foundation of knowledge to guide behaviors in pre-trained model ecosystems. In this work, we present the first empirical investigation of PTM reuse. We interviewed 12 practitioners from the most popular PTM ecosystem, Hugging Face, to learn the practices and challenges of PTM reuse. From this data, we model the decision-making process for PTM reuse. Based on the identified practices, we describe useful attributes for model reuse, including provenance, reproducibility, and portability. Three challenges for PTM reuse are missing attributes, discrepancies between claimed and actual performance, and model risks. We substantiate these identified challenges with systematic measurements in the Hugging Face ecosystem. Our work informs future directions on optimizing deep learning ecosystems by automated measuring useful attributes and potential attacks, and envision future research on infrastructure and standardization for model registries.