Recent theories suggest that Neural Scaling Laws arise whenever the task is linearly decomposed into units that are power-law distributed. Alternatively, scaling laws also emerge when data exhibit a hierarchically compositional structure, as is thought to occur in language and images. To unify these views, we consider classification and next-token prediction tasks based on probabilistic context-free grammars—probabilistic models that generate data via a hierarchy of production rules. For classification, we show that having power-law distributed production rules results in a power-law learning curve with an exponent depending on the rules’ distribution and a large multiplicative constant that depends on the hierarchical structure. By contrast, for next-token prediction, the distribution of production rules controls the fine details of the learning curve, but not the exponent describing the large-scale behaviour.

Understanding representational similarity between neural recordings and computational models is essential for neuroscience, yet remains challenging to measure reliably due to the constraints on the number of neurons that can be recorded simultaneously. In this work, we apply tools from Random Matrix Theory to investigate how such limitations affect similarity measures, focusing on Centered Kernel Alignment (CKA) and Canonical Correlation Analysis (CCA). We propose an analytical framework for representational similarity analysis that relates measured similarities to the spectral properties of the underlying representations. We demonstrate that neural similarities are systematically underestimated under finite neuron sampling, mainly due to eigenvector delocalization. To overcome this, we introduce a denoising method to infer population-level similarity, enabling accurate analysis even with small neuron samples. Theoretical predictions are validated on synthetic and real datasets, offering practical strategies for interpreting neural data under finite sampling constraints.

Abstract Streamlines are a popular method of choice in many flow visualization techniques due to their simplicity and intuitiveness. This paper presents a novel streamline seeding method, which is tailored for visualizing unsteady flow in augmented reality (AR). Our method prioritizes visualizing the visible part of the flow field to enhance the flow representation's quality and reduce the computational cost. Being an image‐based method, it evenly samples 2D seeds from the screen space. Then, a ray is fired toward each 2D seed, and the on‐the‐ray point, which has the largest entropy, is selected. It is taken as the 3D seed for a streamline. By advecting such 3D seeds in the velocity field, which is continuously updated in real time, the unsteady flow is visualized more naturally, and the temporal coherence is achieved with no extra efforts. Our method is tested using an AR application for visualizing airflow from a virtual air conditioner. Comparison with the baseline methods shows that our method is suitable for visualizing unsteady flow in AR.

Abstract This paper proposes a novel method for simulating hyperelastic solids with Smoothed Particle Hydrodynamics (SPH). The proposed method extends the coverage of the state‐of‐the‐art elastic SPH solid method to include different types of hyperelastic materials, such as the Neo‐Hookean and the St. Venant‐Kirchoff models. To this end, we reformulate an implicit integration scheme for SPH elastic solids into an optimization problem and solve the problem using a general‐purpose quasi‐Newton method. Our experiments show that the Limited‐memory BFGS (L‐BFGS) algorithm can be employed to efficiently solve our optimization problem in the SPH framework and demonstrate its stable and efficient simulations for complex materials in the SPH framework. Thanks to the nature of our unified representation for both solids and fluids, the SPH formulation simplifies coupling between different materials and handling collisions.

Abstract Informal communications media pose new challenges for information‐systems design, but the nature of informal interaction offers new opportunities as well. This paper describes NetLens‐E‐mail, a system designed to support exploration of the content‐actor network in large e‐mail collections. Unique features of NetLens‐E‐mail include close coupling of orientation, specification, restriction, and expansion, and introduction and incorporation of a novel capability for iterative projection between content and actor networks within the same collection. Scenarios are presented to illustrate the intended employment of NetLens‐E‐mail, and design walkthroughs with two domain experts provide an initial basis for assessment of the suitability of the design by scholars and analysts.

Databases often contain uncertain and imprecise references to real-world entities. Entity resolution, the process of reconciling multiple references to underlying real-world entities, is an important data cleaning process required before accurate visualization or analysis of the data is possible. In many cases, in addition to noisy data describing entities, there is data describing the relationships among the entities. This relational data is important during the entity resolution process; it is useful both for the algorithms which determine likely database references to be resolved and for visual analytic tools which support the entity resolution process. In this paper, we introduce a novel user interface, D-Dupe, for interactive entity resolution in relational data. D-Dupe effectively combines relational entity resolution algorithms with a novel network visualization that enables users to make use of an entity's relational context for making resolution decisions. Since resolution decisions often are interdependent, D-Dupe facilitates understanding this complex process through animations which highlight combined inferences and a history mechanism which allows users to inspect chains of resolution decisions. An empirical study with 12 users confirmed the benefits of the relational context visualization on the performance of entity resolution tasks in relational data in terms of time as well as users' confidence and satisfaction.

Due to the growing interest in geospatial data mining and analysis, data cleaning and integration in geospatial data is becoming an important issue. Geospatial entity resolution is the process of reconciling multiple location references to the same real world location within a single data source (deduplication) or across multiple data sources (integration). In this paper, we introduce an interactive tool called GeoDDupe which effectively combines automatic data mining algorithms for geospatial entity resolution with a novel network visualization supporting users' resolution analysis and decisions. We illustrate the GeoDDupe interface with an example geospatial dataset and show how users can efficiently and accurately resolve location entities. Finally, the case study with two real-world geospatial datasets demonstrates the potential of GeoDDupe.

The vision of ubiquitous digital photos has arrived. Yet, despite their broad popularity, significant shortfalls remain in the tools used to manage them. We believe that with a bit more creativity and effort, the photo industry can solve many of these problems, offering tools which better support accurate, rapid, and safe shared annotations with comfortable and efficient browsing and search. In this article, we review a number of projects of ours and others on interfaces for photo management. We describe the problems that we see in existing tools and our vision for improving them.

This paper describes a novel instantiation of a digital photo library in a public access system. It demonstrates how designers can utilize characteristics of a target user community (social constraints, trust, and a lack of anonymity) to provide capabilities that would be impractical in other types of public access systems. It also presents a compact set of design principles and guidelines for ensuring the immediate usability of public access information systems. These principles and guidelines were derived from our experience developing PhotoFinder Kiosk, a community photo library. Attendees of a major HCI conference (CHI 2001 Conference on Human Factors in Computing Systems) successfully used the tool to browse and annotate collections of photographs spanning 20 years of HCI-related conferences, producing a richly annotated photo history of the field of human-computer interaction. Observations and log data were used to evaluate the tool and develop the guidelines. They provide specific guidance for practitioners, as well as a useful framework for additional research in public access interfaces.

Annotating photographs is such a time-consuming, tedious and error-prone data entry task that it discourages most owners of personal photo libraries. By allowing the user to drag labels, such as personal names, from a scrolling list and drop them onto a photo, we believe we can make the task faster, easier and more appealing. Since the names are entered in a database, searching for all photos of a friend or family member is dramatically simplified. We describe the user interface design and the database schema to support direct annotation, as implemented in our PhotoFinder prototype.