Portland Community College

Neural network representations are often analyzed as vectors in a fixed Euclidean space. However, their coordinates are not uniquely defined. If a hidden representation is transformed by an invertible linear map, the network function can be preserved by applying the inverse transformation to downstream weights. Representations are therefore defined only up to invertible linear transformations. We study neural representation spaces from this geometric viewpoint and treat them as vector spaces with a gauge freedom under the general linear group. Within this framework, commonly used similarity measures such as cosine similarity become metric-dependent quantities whose values can change under coordinate transformations that leave the model function unchanged. This provides a common interpretation for several observations in the literature, including cosine-similarity instability, anisotropy in embedding spaces, and the appeal of representation comparison methods such as SVCCA and CKA. Experiments on multilayer perceptrons and convolutional networks confirm that inserting invertible transformations into trained models can substantially distort cosine similarity and nearest-neighbor structure while leaving predictions unchanged. These results indicate that analysis of neural representations should focus either on quantities that are invariant under this gauge freedom or on explicitly chosen canonical coordinates.

Insider threat detection is difficult because malicious behavior is rare, irregular, and buried in long periods of inactivity. In enterprise audit data, most windows contain little activity, while attacks appear intermittently and range from brief events to sustained campaigns. Standard reconstruction-based models are therefore dominated by inactive regions and tend to learn baseline behavior rather than meaningful deviations. We separate activity presence from magnitude. Each window is decomposed into a binary mask indicating whether activity occurs and a value matrix capturing its intensity. A dual-channel autoencoder reconstructs both, with value loss applied only where activity is present, directing learning toward sparse structure. Using the CERT r5.2 dataset as a controlled setting, we examine how anomaly signal changes with temporal configuration. Short attacks are detected mainly through presence; longer attacks introduce a magnitude component; noise degrades magnitude reliability and shifts detection back toward presence. The balance between channels is not fixed and follows the data. At the campaign level, signal concentrates in a small number of anomalous windows. Simple aggregation that emphasizes extreme scores is sufficient to recover extended activity without explicit sequence modeling. Effective detection depends less on model complexity and more on aligning representation and objective with sparse temporal structure.

Environmental effects such as hierarchical triple motion can introduce cumulative phase modulation in gravitational-wave signals through time-dependent line-of-sight acceleration. Whether such smooth time-warp distortions are observable depends jointly on deformation strength and signal-to-noise ratio (SNR), yet this relationship has not been quantified in a template-free framework. We study the detectability of these distortions using time-frequency representations derived from the continuous wavelet transform. Instead of reconstruction error alone, we analyze trajectory-level statistics, in particular the evolution of the power-weighted frequency centroid. We find that environmental modulation can be detected using a single-sample statistic referenced to an isolated-binary distribution, without requiring matched templates. Across a grid of cumulative phase distortions and SNR, detection performance collapses onto a single scaling parameter defined as Lambda = Delta phi x SNR. The ROC-AUC follows an approximately sigmoid transition in this parameter. Moderate distortions are detectable even at low SNR, whereas smaller distortions require higher SNR. These results indicate that smooth environmental phase modulation is not generically absorbed by intrinsic waveform variability; instead, detectability is governed by a simple scaling between cumulative phase distortion and signal strength.

IntroductionIn 2014, the National Institutes of Health (NIH) invested in the Building Infrastructure Leading to Diversity (BUILD) initiative to enhance diversity in the biomedical research workforce. As one of ten grantees nationwide, the BUILD EXITO project at Portland State University established an institutionally and geographically diverse consortium including local community colleges, a research-intensive medical institution, and universities and community colleges around the Pacific Rim. The goal of this collaboration was to support comprehensive research training for undergraduates from backgrounds historically underrepresented in the biomedical workforce. This manuscript aims to provide insights into creating and sustaining a large-scale multi-institutional consortium.MethodsUsing a collaborative and reflective approach, this study presents a collective account of developing and sustaining a decade-long equity-focused partnership. The authors, all deeply involved in the partnership, participated in a series of semi-structured conversations designed to elicit strategies and lessons learned for building and sustaining multi-institutional collaborations.ResultsThree main themes arose from the reflections on core strategies for creating and maintaining the partnership: 1) having a robust framework for diverse, equitable, and inclusive partnership, 2) equitable, flexible opportunities for goal setting and program implementation, and 3) planning for sustainability from the beginning. Obstacles faced throughout the decade-long partnership include the retention of all partners and the tension between institutional buy-in and the pursuit of external funding. Finally, the Partners defined two lessons learned from the EXITO experience: 1) the importance of a critical mass of stakeholders, and 2) the need to expand institutional leadership teams for partner sustainability.DiscussionWhile working across institutional boundaries may present challenges, multi-institutional partnerships allow for a broader reach to diverse student populations and create meaningful access to opportunities that may not otherwise exist. The EXITO infrastructure serves as a model for developing and sustaining partnerships for equity-focused student programs.