Global ETD Search

Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.

Results

Showing 1 to 17 of 17 for “"Low rank factorization"”.

  1. Compression and acceleration of machine learning algorithms based on low rank factorization

    … frameworks for effectively applying existing low-rank factorization (LRF) and pruning techniques in real-world DNN deployment scenarios. All methodologies operate entirely in the post-training phase, requiring no retraining from scratch and only minimal fine-tuning. LRF decomposes weight …

    greece Repository record for Compression and acceleration of machine learning algorithms based on low rank factorization (opens in a new tab)

  2. Dimensionality-reduced estimation of primaries by sparse inversion

    … is approximated adaptively by a randomized low-rank factorization. Compared to conventional methods, which need passes through all the data possibly including on-the-fly interpolations for each iteration, our approach has the advantage that the passes are reduced to one to three. In …

    ubc Repository record for Dimensionality-reduced estimation of primaries by sparse inversion (opens in a new tab) Download PDF for Dimensionality-reduced estimation of primaries by sparse inversion (opens in a new tab)

  3. Heuristics for large-scale semidefiniite programming for the K disjoint clique problem

    … problem as a SDP based on non-convex low rank factorization, and solve using Alternating Direction Method of Multipliers, augmented Lagrangian, and alternating direction. We will present numerical results illustrating the efficacy of our approach for clustering of real and simulated …

    alabama Repository record for Heuristics for large-scale semidefiniite programming for the K disjoint clique problem (opens in a new tab)

  4. Privacy-aware publication and utilization of healthcare data

    … graphical models, non-parametric modeling, and low-rank factorization techniques. It shows that much utility can be extracted from health records while maintaining strong privacy guarantees and protection of sensitive health information.

    texas Repository record for Privacy-aware publication and utilization of healthcare data (opens in a new tab)

  5. Reconstructing dynamic correlation and bonding in curved pi-systems

    … advances, highlighting how representation: low-rank factorization on the methods side and localized chemically intuitive orbitals on the analysis side, enables both scalable computation and clearer interpretation, and by outlining opportunities for GPU kernels.

    iastate Repository record for Reconstructing dynamic correlation and bonding in curved pi-systems (opens in a new tab)

  6. A multiresolution approach for tensor completion from coarse and partial observations

    … recover the original fine-granular tensor with low-rank factorization. We formulate a coupled tensor completion problem and propose an efficient Multi-resolution Tensor Completion model (MTC) to solve the problem. Our MTC model explores tensor mode properties and leverages the hierarchy of …

    uiuc Repository record for A multiresolution approach for tensor completion from coarse and partial observations (opens in a new tab)

  7. Low rank methods for optimizing clustering

    … often have the majority of information in a low rank subspace. By careful exploitation of these low rank structures in clustering problems, we find new optimization approaches that reduce the memory and computational cost. We discuss two cases where this arises. First, we consider the …

    purdue-thes Repository record for Low rank methods for optimizing clustering (opens in a new tab)

  8. Matrix free methods for large scale optimization

    … definiteness of covariance matrices for general rank 2 updates. An active set strategy is employed to speed up BCD for hub structure problem. For latent variable selection problems, we propose a method for maintaining a low rank factorization for the covariance matrix while preserving the …

    washington Repository record for Matrix free methods for large scale optimization (opens in a new tab)

  9. Advances in Bayesian Latent Factor Interaction Models for Ecological and Environmental Health Applications

    … bias due to uncertain species occurrence. This allows non-edges arising from species non-overlap to be more effectively distinguished from true absences of interaction.</p><p>Second, we consider spatiotemporal populations of highly sparse networks, where both node set variability and …

    duke Repository record for Advances in Bayesian Latent Factor Interaction Models for Ecological and Environmental Health Applications (opens in a new tab)

  10. Order reduction of large thermal and electrical models with system-theoretic techniques and matrix equation algorithms

    … eliminate computational costs and storage needs, low-rank factorization methods have been developed that allow the use of these methods in real-world applications. The emphasis in the present work is given in circuit and thermal models derived from integrated circuits and the calculation of the …

    greece Repository record for Order reduction of large thermal and electrical models with system-theoretic techniques and matrix equation algorithms (opens in a new tab)

  11. Tensor-Network Simulation of Quantum Spin Systems with Bond-Dependent Interactions

    … larger or even infinite-sized systems through a low-rank factorization of the state coefficients. Tensor operations are typically highly parallelizable, and therefore capable of utilizing the high performance of modern general-purpose graphics processing units (GPU). A particular tensor-network …

    mcmaster Repository record for Tensor-Network Simulation of Quantum Spin Systems with Bond-Dependent Interactions (opens in a new tab)

  12. Responsible Recklessness: Why High Performance Computing Must Embrace Randomized Numerical Linear Algebra

    … for the two NLA problem classes: full matrix factorizations through QR decomposition with column pivoting (QRCP), and low-rank factorization via Singular Value Decomposition (SVD). The core contribution of the projects I present lies in my development of high-performance implementations for …

    utk-theses Repository record for Responsible Recklessness: Why High Performance Computing Must Embrace Randomized Numerical Linear Algebra (opens in a new tab)

  13. Active and Adaptive Techniques for Learning over Large-Scale Graphs

    … task. As node embedding is naturally viewed as a low-rank factorization of a node-to-node similarity matrix, a versatile approach is introduced to learn the similarity matrix of a given graph with minimal computational overhead, and in a fully unsupervised manner. Extensive experimentation using …

    umn Repository record for Active and Adaptive Techniques for Learning over Large-Scale Graphs (opens in a new tab)

  14. Vector Quantization of Deep Convolutional Neural Networks with Learned Codebook

    … compression, including pruning, quantization, low-rank factorization and knowledge distillation approaches. Our main interest is quantization techniques, which compress networks by reducing the precision of parameters. Full-precision weights, activations and even gradients in networks can be …

    ottawa-retro Repository record for Vector Quantization of Deep Convolutional Neural Networks with Learned Codebook (opens in a new tab)

  15. Quantum Algorithms for Scientific Computation

    … formula, a high-order method constructed from low-order product formulas. First, it provides a detailed characterization of the error scaling of the time-independent multiproduct formula for simulating quantum dynamics in finite-dimensional quantum systems. This analysis establishes both …

    maryland Repository record for Quantum Algorithms for Scientific Computation (opens in a new tab)

  16. Using global low-rank kernel matrix approximations in machine learning and uncertainty quantification

    … which uses randomized sampling to construct a low-rank factorization of the matrix. We implement a parallel version of Nystrom and use it to study properties of large kernel matrices and offer both theoretical and empirical comparisons of Nystrom and treecodes (a more sophisticated matrix …

    texas Repository record for Using global low-rank kernel matrix approximations in machine learning and uncertainty quantification (opens in a new tab)

  17. Scalable and Efficient Training of Foundation Models

    … is computational and memory efficiency. While low-rank adaptation (LoRA) can be utilized to alleviate the computational overhead, it suffers from slow optimization convergence and notable performance drop. Chapter 2 analyzes an often overlooked perspective, namely residual topology, in model …

    umn Repository record for Scalable and Efficient Training of Foundation Models (opens in a new tab)