Wearable e-textile interfaces require gesture recognition capabilities but face severe constraints in power consumption, computational capacity, and form factor that make traditional deep learning impractical. While lightweight architectures like MobileNet improve efficiency, they still demand thousands of parameters, limiting deployment on textile-integrated platforms. We introduce a convexified attention mechanism for wearable applications that dynamically weights features while preserving convexity through nonexpansive simplex projection and convex loss functions. Unlike conventional attention mechanisms using non-convex softmax operations, our approach employs Euclidean projection onto the probability simplex combined with multi-class hinge loss, ensuring global convergence guarantees. Implemented on a textile-based capacitive sensor with four connection points, our approach achieves 100.00% accuracy on tap gestures and 100.00% on swipe gestures—consistent across 10-fold cross-validation and held-out test evaluation—while requiring only 120–360 parameters, a 97% reduction compared to conventional approaches. With sub-millisecond inference times (290–296 μ s) and minimal storage requirements (< 7KB), our method enables gesture interfaces directly within e-textiles without external processing. Our evaluation, conducted in controlled laboratory conditions with a single-user dataset, demonstrates feasibility for basic gesture interactions. Real-world deployment would require validation across multiple users, environmental conditions, and more complex gesture vocabularies. These results demonstrate how convex optimization can enable efficient on-device machine learning for textile interfaces.

The increasing availability of big data and adoption of sophisticated computational techniques in biomedical research has exciting implications for our scientific understanding of human health. However, researchers report struggling to find data science education that meets their needs, despite the fact that many training programs and online resources exist. There is a lack of evidence on the strengths and weaknesses of various training options, making selecting an educational path daunting. We created a new data science training program focused on rigorous, reproducible methods for biomedical research, making use of tightly scoped modular content that can be flexibly arranged to provide a curriculum tailored to a researcher's specific needs and skill level. Moreover, we ran a study testing the program's effectiveness, providing not only another option for data science training but also a model for collecting and sharing relevant data on data science education programs. We ran two waves of research participants, adjusting our materials in between to improve both the training program and our research design. For both waves, we pre-registered hypotheses that learners' self-reported data science ability and level of agreement with important tenets of open science would increase over the course of the program. Indeed, learners showed significant improvement in data science ability (Wave 1: t(47) = 10.18, p < .001, Wave 2: t(238) = 17.12, p < .001) and greater agreement with open science values (Wave 1: t(47) = 3.56, p < .001, Wave 2: t(238) = 7.95, p < .001). Follow up analyses underscore the robustness of improvement in data science ability. The improvement in open science values was more moderate and was significant only in some of our pre-registered hypothesis tests, likely due to a ceiling effect as most learners reported high agreement with open science values at pretest.

Recent work has investigated the construction of touch-sensitive knitted fabrics, capable of being manufactured at scale, and having only two connections to external hardware. Additionally, several sensor design patterns and application prototypes have been introduced. Our aim is to start shaping the future of this technology according to user expectations. Through a formative focus group study, we explore users’ views of using these fabrics in different contexts and discuss potential concerns and application areas. Subsequently, we take steps toward addressing relevant questions, by first providing design guidelines for application designers. Furthermore, in one user study, we demonstrate that it is possible to distinguish different swipe gestures and identify accidental contact with the sensor, a common occurrence in everyday life. We then present experiments investigating the effect of stretching and laundering of the sensors on their resistance, providing insights about considerations necessary to include in computational models.

Mesh is a preeminent and efficient data structure for 3D objects that support high-resolution representation. Recent deep learning techniques applied to unstructured mesh data define rigid convolutions that fail to capture the rich geometric and topological attributes of the mesh. We propose an efficient, deterministic process that converts a mesh into a hybrid graph and captures the geometric features of its constituting components: vertices, edges, and faces. In addition, we introduce a novel representation learning framework that encodes mesh elements by focusing on the most relevant parts of the geometric structure using a dual-level attention architecture. We evaluate the efficacy of the proposed representation in the context of the 3D shape segmentation problem. The superior performance of the proposed representation to the state of the art in supervised segmentation illustrates the soundness of the proposed attention model.

In a world becoming dependent on cloud computing, multiple and many techniques are needed to meet security and privacy requirements. This study investigates using Fully Homomorphic Encryption (FHE) in the cloud for a set of signal processing algorithms. Using FHE provides a level of security and privacy for cloud computing. This research implements a framework with a set of real number processing formats that allows FHE to be used in the cloud. Developing formats for real numbers is required because FHE operates over integers. Both a fixed point binary format and a rational number format are implemented to allow for real number processing. Experimentally, the framework is implemented and verified with three algorithms: fast Fourier transform, speeded up robust features, and histogram of oriented gradients. Each implementation is bounded with error caused by the formats and examined against images (two‐dimensional signals).

Fabric sensors have been introduced to enable flexible touch-based interaction. We advance the technical capabilities of a scalable and low-profile knitted capacitive touch sensing system by introducing methods to improve its touch localization accuracy. The sensor hardware design tends toward minimalism by using a single conductive yarn and two external connections located at each endpoint. Fewer connectors simplify the textile system integration, but this comes at the expense of reduced signal information output from the system. The electrical continuity of the sensing element, essential to the process of knitting, also increases the uncertainty of localizing touch. We propose using Bode analysis to measure changes in signal due to capacitive touch, as well as design a new algorithm, MSD, which retains the most significant aspects of the signal in terms of touch location identification. We do not classify location of touch, but focus on an invariant signal representation. To evaluate our methods, we introduce ELD, a distance metric to compute the similarity of pairs of key-presses, generalizable to computing distances of tensors of varying lengths. Our experiments show that the proposed sensing method results in high-fidelity signals. Furthermore, the sparse representation of key-presses produced by MSD significantly increases separability between different touch locations. Possible applications based on these sensors are also illustrated through prototypes and use case descriptions.

It has been argued by Thom and Palm that sparsely-connected neural networks (SCNs) show improved performance over fully-connected networks (FCNs). Super-regular networks (SRNs) are neural networks composed of a set of stacked sparse layers of (epsilon, delta)-super-regular pairs, and randomly permuted node order. Using the Blow-up Lemma, we prove that as a result of the individual super-regularity of each pair of layers, SRNs guarantee a number of properties that make them suitable replacements for FCNs for many tasks. These guarantees include edge uniformity across all large-enough subsets, minimum node in-and out-degree, input-output sensitivity, and the ability to embed pre-trained constructs. Indeed, SRNs have the capacity to act like FCNs, and eliminate the need for costly regularization schemes like Dropout. We show that SRNs perform similarly to X-Nets via readily reproducible experiments, and offer far greater guarantees and control over network structure.

Labeling problems are finding increasing applications to optimization problems. They usually get realized into linear or quadratic optimization problems, which are inefficient for large graphs. In this paper we propose an efficient primal-dual solution, MLPD, for a family of labeling problems. We apply this algorithm to the analysis of immune repertoires, and compare it against our baseline approach based on refinement operators. We provide a comparative evaluation both in terms of accuracy and computational efficiency with respect to the baseline model, as well as to quadratic optimization.

We study downlink delay minimization within the context of cellular user association policies that map mobile users to base stations. We note the delay minimum user association problem fits within a broader class of network utility maximization and can be posed as a non-convex quadratic program. This non-convexity motivates a split quadratic objective function that captures the original problem's inherent tradeoff: association with a station that provides the highest signal-to-interference-plus-noise ratio (SINR) vs. a station that is least congested. We find the split-term formulation is amenable to linearization by embedding the base stations in a hierarchically well-separated tree (HST), which offers a linear approximation with constant distortion. We provide a numerical comparison of several problem formulations and find that with appropriate optimization parameter selection, the quadratic reformulation produces association policies with sum delays that are close to that of the original network utility maximization. We also comment on the more difficult problem when idle base stations (those without associated users) are deactivated.

Music classification is a challenging problem with many applications in today's large-scale datasets with Gigabytes of music files and associated metadata and online streaming services. Recent success with deep neural network architectures on large-scale datasets has inspired numerous studies in the machine learning community for various pattern recognition and classification tasks such as automatic speech recognition, natural language processing, audio classification and computer vision. In this paper, we explore a two-layer neural network with manifold learning techniques for music genre classification. We compare the classification accuracy rate of deep neural networks with a set of well-known learning models including support vector machines (SVM and '1-SVM), logistic regression and '1-regression in combination with hand-crafted audio features for a genre classification task on a public dataset. Our experimental results show that neural networks are comparable with classic learning models when the data is represented in a rich feature space.

Performing signal processing via a distributed method while maintaining data privacy has great uses. We focus on this problem to a two party situation where the first one (client) has a signal needing to be processed and is computationally bounded and the second (server) has computational resources. As a client, revealing the signal unencrypted causes a violation of privacy. One solution to this problem is to process the signal while encrypted. Problems of this type have been attracting attention recently; particularly with the growing capabilities of cloud computing. This paper contributes to solving this type of problem by processing the signals in an encrypted form, using fully homomorphic encryption (FHE). Realization of this type of application was performed using a brightness/contrast filter as a simple form of signal processing. Three additional contributions of this manuscript includes (1) extending FHE to real numbers, (2) bounding the error related to the FHE process against the unencrypted variation of the process, and (3) increasing the practicality of FHE as a tool by using graphical processing units (GPU).

As the demand for multimedia grows, the development of information retrieval systems utilizing all available data modalities becomes of paramount importance. The provision of multiple modalities is motivated by usability, presence of noise in one modality and non-universality of a single modality. Radio stations and music TV channels hold archives of millions of music tapes and lyrics. Gigabytes of music files are also spread over the web along with the lyrics and metadata for each file. Searching and organizing large scale multimodal datasets is a challenging task. Supervised methods such as support vector machine (SVM) achieve state of the art performance for music classification on single modality, but suffer from over-fitting on training examples and limitations of single modality approaches. In this paper, we introduce a classifier fusion of multimodal audio and lyrics data to address these single modality classification limitations. We introduce the multimodal l1-SVM classifier, that utilizes sparse methods to deal with over-fitting for music classification. We compare the classification accuracy of the fusion classifier for a genre classification task in a large public dataset with single modality l1-SVM.

Humans interact with each other using different communication modalities including speech, gestures and written documents. In the absence of one modality or presence of a noisy modality, other modalities can benefit precision of systems. HCI systems can also benefit from these multimodal communication models for different machine learning tasks. The provision of multiple modalities is motivated by usability, presence of noise in one modality and non-universality of a single modality. Combining multimodal information introduces new challenges to machine learning such as designing fusion classifiers. In this paper we explore the multimodal fusion of audio and lyrics for music artist identification. We compare our results with a single modality artist classifier and introduce new directions for designing a fusion classifier.

In this paper, we propose an efficient embedding for modeling higher-order (n-gram) phrases that projects the n-grams to low-dimensional latent semantic space, where a classification function can be defined. We utilize a deep neural network to build a unified discriminative framework that allows for estimating the parameters of the latent space as well as the classification function with a bias for the target classification task at hand. We apply the framework to large-scale sentimental classification task. We present comparative evaluation of the proposed method on two (large) benchmark data sets for online product reviews. The proposed method achieves superior performance in comparison to the state of the art.

In this work we propose a novel variational method that we intend to use for estimating non-rigid texture deformation. The method is able to capture variation in gray scale images with respect to the geometry of its features. Accurate localization of features in the presence of unknown deformations is a crucial property for texture characterization. Our experimental evaluations demonstrate that accounting for geometry of features in texture images leads to significant improvements in localization of these features, when textures undergo geometrical transformations. In addition, feature descriptors using geometrical total variation energies discriminate between various regular textures with accuracy comparable to SIFT descriptors, while reduced dimensionality of TVG descriptor yields significant improvements over SIFT in terms of retrieval time.

A common problem in many areas of behavioral research is the analysis of the large volume of data recorded during the execution of the tasks being studied. Recent work has proposed the use of an automated method based on canonical sets to identify the most representative patterns in a large data set, and described an initial experiment in identifying canonical web-browsing patterns. However, there is a significant limitation to the method: it requires the similarity matrix to be symmetric, and thus can only be used for problems that can be modeled as unoriented topologies. In this paper we propose a novel enhancement to the method to support oriented topologies by allowing the similarity matrix to be nonsymmetric. We demonstrate the power of this new technique by applying the new method to find canonical lane changes in a driving simulator experiment.

While the collection of behavioral protocols has been common practice in human-computer interaction research for many years, the analysis of large protocol data sets is often extremely tedious and time-consuming, and automated analysis methods have been slow to develop. This paper proposes an automated method of protocol analysis to find canonical behaviors --- a small subset of protocols that is most representative of the full data set, providing a reasonable "big picture" view of the data with as few protocols as possible. The automated method takes advantage of recent algorithmic developments in computational vision, modifying them to allow for distance measures between behavioral protocols. The paper includes an application of the method to web-browsing protocols, showing how the canonical behaviors found by the method match well to sets of behaviors identified by expert human coders.

Learning a class prototype from a set of exemplars is an important challenge facing researchers in object categorization. Although the problem is receiving growing interest, most approaches assume a one-to-one correspondence among local features, restricting their ability to learn true abstractions of a shape. In this paper, we present a new technique for learning an abstract shape prototype from a set of exemplars whose features are in many-to-many correspondence. Focusing on the domain of 2D shape, we represent a silhouette as a medial axis graph whose nodes correspond to "parts" defined by medial branches and whose edges connect adjacent parts. Given a pair of medial axis graphs, we establish a many-to-many correspondence between their nodes to find correspondences among articulating parts. Based on these correspondences, we recover the abstracted medial axis graph along with the positional and radial attributes associated with its nodes. We evaluate the abstracted prototypes in the context of a recognition task.

Managers are often unable to explain objectively why or when effort was misplaced during the development process. In this paper, we present a formal technique to depict the expended effort during the life-cycle of a software feature using feature development manifolds (FDMs). Using the FDMs we can compute the preferred development path for a given feature. This development path includes the versions of a software feature that contributed to the final version of the feature in a positive way. The preferred development path excludes versions of the software feature that should have been skipped. Once the preferred development path is computed the amount of wasted effort can be quantified using the metric that we have developed. We demonstrate the effectiveness of our approach to compute wasted software feature development by applying our technique to two large open source software systems, Gaim and Firefox.

The implementations of software features evolve as an application matures. We define a measure of feature implementation overlap that determines how similar features are in their execution by examining their call graphs. We consider how this measure changes over time, and evaluate the hypothesis that over time and subsequent versions of a software application, the implementations of semantically similar features converge. As the features of an application converge in their implementation, we are able to more effectively determine groups of semantically similar features and to reduce the cost of program comprehension by selecting few key features that give an overview of the system. We present a case study analyzing the features of the Jext, Firefox, and Gaim software systems to support our hypothesis.

In this paper, we present an approach that examines the evolution of code stored in source control repositories. The technique identifies change clusters, which can help managers to classify different code change activities as either software maintenance or a new development. Furthermore, identifying the variations in change clusters over time exposes trends in the development of a software system. We present a case study that uses a sequence of change clusters to track the evolution of the PostgreSQL software project. Our case study demonstrates that our technique reveals interesting patterns about the progress of code development within each release of PostgreSQL. We show that the increase in the number of clusters not only identifies the areas where development has occurred, but also reflects the amount of structural change in code. We also compare how the change clusters vary over time in order to make generalizations about the focus of development

Recent work in the object recognition community has yielded a class of interest-point-based features that are stable under significant changes in scale, viewpoint, and illumination, making them ideally suited to landmark-based navigation. Although many such features may be visible in a given view of the robot's environment, only a few such features are necessary to estimate the robot's position and orientation. In this paper, we address the problem of automatically selecting, from the entire set of features visible in the robot's environment, the minimum (optimal) set by which the robot can navigate its environment. Specifically, we decompose the world into a small number of maximally sized regions, such that at each position in a given region, the same small set of features is visible. We introduce a novel graph theoretic formulation of the problem, and prove that it is NP-complete. Next, we introduce a number of approximation algorithms and evaluate them on both synthetic and real data. Finally, we use the decompositions from the real image data to measure the localization performance versus the undecomposed map

Typical documentation for object-oriented programs includes descriptions of the parameters and return types of each method in a class, but little or no information on valid method invocation sequences. Knowing the sequence with which methods of a class can be invoked is useful information especially for software engineers (e.g., developers, testers) who are actively involved in the maintenance of large software systems. This paper describes a new approach and a tool for generating class usage scenarios (i.e., how a class is used by other classes) from method invocations, which are collected during the execution of the software. Our approach is algorithmic and employs the notion of canonical sets to categorize method sequences into groups of similar sequences, where each group represents a usage scenario for a given class.

We present a polynomial-time algorithm for a special case of the Euclidean traveling salesman problem in which a robot must visit all the vertices of two non-intersecting polygons without crossing any polygon edge. If both polygons are convex, one enclosing the other, our algorithm can find the optimal tour of the channel between them in time O(m/sup 3/ + m/sup 2/n) and O(nm + m/sup 2/) space, where the exterior polygon has n vertices and the interior m vertices. In the more general case of non-convex polygons (not necessarily nested), the algorithm finds the exact optimum tour in O(n/sup 2/m + m/sup 3/) time and O(n/sup 2/ + m/sup 2/) space. At the end we give several examples in the context of robot navigation.

We present a 3D matching framework based on a many-to-many matching algorithm that works with skeletal representations of 3D volumetric objects. We demonstrate the performance of this approach on a large database of 3D objects containing more than 1000 exemplars. The method is especially suited to matching objects with distinct part structure and is invariant to part articulation. Skeletal matching has an intuitive quality that helps in defining the search and visualizing the results. In particular, the matching algorithm produces a direct correspondence between two skeletons and their parts, which can be used for registration and juxtaposition.

Hierarchical image structures are abundant in computer vision and have been used to encode part structure, scale spaces, and a variety of multiresolution features. In this paper, we describe a framework for indexing such representations that embeds the topological structure of a directed acyclic graph (DAG) into a low-dimensional vector space. Based on a novel spectral characterization of a DAG, this topological signature allows us to efficiently retrieve a promising set of candidates from a database of models using a simple nearest-neighbor search. We establish the insensitivity of the signature to minor perturbation of graph structure due to noise, occlusion, or node split/merge. To accommodate large-scale occlusion, the DAG rooted at each nonleaf node of the query "votes" for model objects that share that "part," effectively accumulating local evidence in a model DAG's topological subspaces. We demonstrate the approach with a series of indexing experiments in the domain of view-based 3D object recognition using shock graphs.

Recent work in the object recognition community has yielded a class of interest point-based features that are stable under significant changes in scale, viewpoint, and illumination, making them ideally suited to landmark-based navigation. Although many such features may be visible in a given view of the robot's environment, only a few such features are necessary to estimate the robot's position and orientation. In this paper, we address the problem of automatically selecting, from the entire set of features visible in the robot's environment, the minimum (optimal) set by which the robot can navigate its environment. Specifically, we decompose the world into a small number of maximally sized regions such that at each position in a given region, the same small set of features is visible. We introduce a novel graph theoretic formulation of the problem and prove that it is NP-complete. Next, we introduce a number of approximation algorithms and evaluate them on both synthetic and real data.

Graph matching is an important component in many object recognition algorithms. Although most graph matching algorithms seek a one-to-one correspondence between nodes, it is often the case that a more meaningful correspondence exists between a cluster of nodes in one graph and a cluster of nodes in the other. We present a matching algorithm that establishes many-to-many correspondences between nodes of noisy, vertex-labeled weighted graphs. The algorithm is based on recent developments in efficient low-distortion metric embedding of graphs into normed vector spaces. By embedding weighted graphs into normed vector spaces, we reduce the problem of many-to-many graph matching to that of computing a distribution-based distance measure between graph embeddings. We use a specific measure, the earth mover's distance, to compute distances between sets of weighted vectors. Empirical evaluation of the algorithm on an extensive set of recognition trials demonstrates both the robustness and efficiency of the overall approach.

In an object recognition system, if the extracted image features are multilevel or multiscale, the indexing structure may take the form of a tree. Such structures are not only common in computer vision, but also appear in linguistics, graphics, computational biology, and a wide range of other domains. In this paper, we develop an indexing mechanism that maps the topological structure of a tree into a low-dimensional vector space. Based on a novel eigenvalue characterization of a tree, this topological signature allows us to efficiently retrieve a small set of candidates from a database of models. To accommodate occlusion and local deformation, local evidence is accumulated in each of the tree's topological subspaces. We demonstrate the approach with a series of indexing experiments in the domain of 2-D object recognition.

We introduce a novel view-based object representation, called the saliency map graph (SMG), which captures the salient regions of an object view at multiple scales using a wavelet transform. This compact representation is highly invariant to translation, rotation (image and depth), and scaling, and offers the locality of representation required for occluded object recognition. To compare two saliency map graphs, we introduce two graph similarity algorithms. The first computes the topological similarity between two SMG's, providing a coarse-level matching of two graphs. The second computes the geometrical similarity between two SMG's, providing a fine-level matching of two graphs. We test and compare these two algorithms on a large database of model object views.

We have been developing a theory for the generic representation of 2-D shape, where structural descriptions are derived from the shocks (singularities) of a curve evolution process, acting on bounding contours. We now apply the theory to the problem of shape matching. The shocks are organized into a directed, acyclic shock graph, and complexity is managed by attending to the most significant (central) shape components first. The space of all such graphs is highly structured and can be characterized by the rules of a shock graph grammar. The grammar permits a reduction of a shockgraph to a unique rooted shock tree. We introduce a novel tree matching algorithm which finds the best set of corresponding nodes between two shock trees in polynomial time. Using a diverse database of shapes, we demonstrate our system's performance under articulation, occlusion, and changes in viewpoint.