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 11 of 11 for “"landscape geometry"”.

  1. Optimization and Loss Landscape Geometry of Deep Learning

    … We achieve this by studying the loss landscape geometry of deep neural networks. This is extremely challenging due to non-convex objective functions and extremely high-dimensional parameter spaces. We address this by first studying a simple class of neural networks: two-layer linear …

    toronto-retro Repository record for Optimization and Loss Landscape Geometry of Deep Learning (opens in a new tab)

  2. Sub-networks and Spectral Anisotropy in Deep Neural Networks

    … on the roles of implicit regularization, loss landscape geometry, and sparse substructures in modern deep neural networks, offering insights into the geometric nature of DNNs.

    chalmers Repository record for Sub-networks and Spectral Anisotropy in Deep Neural Networks (opens in a new tab)

  3. Insights into neural network learning of structured models : optimization, architectures, and scientific surrogates

    … between neural network architecture, loss landscape geometry, and optimization in scientific machine learning. We analyze optimization of physics-informed neural networks (PINNs), which can exhibit ill-conditioned loss landscapes. We develop a randomized sketching algorithm to approximate …

    ubc Repository record for Insights into neural network learning of structured models : optimization, architectures, and scientific surrogates (opens in a new tab) Download PDF for Insights into neural network learning of structured models : optimization, architectures, and scientific surrogates (opens in a new tab)

  4. A fractal landscape dynamics approach to understanding particle motion in soft jammed materials

    … that soft jammed materials exhibit fractal landscape dynamics, where the particles' motion is not merely random but follows patterns influenced by the system's underlying fractal energy landscape. Through experimental observations, theoretical models, and numerical simulations of ripening …

    penn Repository record for A fractal landscape dynamics approach to understanding particle motion in soft jammed materials (opens in a new tab)

  5. Entropic Approaches to Machine Learning

    … structure of the phenomenon, it injects landscape-geometry specific noise via entropy maximization and allows the appropriate model to emerge from the ensemble. Moreover, simmering's effectiveness indicates that machine “learning” is an emergent phenomenon and that my entropic approaches …

    queens Repository record for Entropic Approaches to Machine Learning (opens in a new tab)

  6. Compound-Gaussian-regularized inverse problems: theory, algorithms, and neural networks

    … to facilitate the learning of the optimization landscape geometry. Third, a generalization on the newly constructed CG regularized least squares iterative algorithm is developed, theoretically analyzed, and unrolled to yield a novel state-of-the-art DNN that provides a partial learning of the …

    colostate Repository record for Compound-Gaussian-regularized inverse problems: theory, algorithms, and neural networks (opens in a new tab)

  7. Understanding the dynamics of tsunamis from signatures in the landscape

    … and the signatures they leave behind in the landscape. Previous sedimentary research has developed methods for identifying tsunami signatures and attempting to distinguish them from related processes, but has not generally provided detailed information about the events that created them. The …

    unsw Repository record for Understanding the dynamics of tsunamis from signatures in the landscape (opens in a new tab)

  8. The Thickness of Landscape, horizontally and vertically considered

    … in Washington, D.C. is a significant industrial landscape that provided safe clean water to much of the capital during 1905-1985. At the outset, the McMillan Commission chose to make the plant and adjacent McMillan reservoir part of the park system it was mapping for Washington. Frederick Law …

    vt Repository record for The Thickness of Landscape, horizontally and vertically considered (opens in a new tab)

  9. Charting the Landscape of Ventral Stream Neural Code on Generative Image Manifolds

    … way. Neural tuning can be conceptualized as a landscape on the natural image manifold, and generative models such as GANs provide a concrete instantiation of the manifolds. We hypothesized that the maximally activating images or peaks on these landscapes are critical for understanding visual …

    wustl Repository record for Charting the Landscape of Ventral Stream Neural Code on Generative Image Manifolds (opens in a new tab)

  10. Exploiting Linear Connectivity modulo Permutation Symmetries in Neural Networks

    … the geometric properties of neural network loss landscapes, focusing on the phenomenon of linear connectivity modulo permutation (LC mod P) symmetries. The initial section of this work formalizes the concept of LC mod P by introducing a clear taxonomy of connectivity: weak, strong, and a new …

    toronto-retro Repository record for Exploiting Linear Connectivity modulo Permutation Symmetries in Neural Networks (opens in a new tab)

  11. Vítejte ve Velvetonu

    Tématem diplomové práce je fikční prostor vizualizovaný s pomocí malířského média. Fikční svět nazvaný Velveton má svou charakteristiku a své vnitřní rozpory. Jako každé prostředí tohoto druhu se i Velveton odvíjí od naší reality, ale funguje jako metaforický svět za zrcadlem. Vzhled tohoto místa …

    brno-tech Repository record for Vítejte ve Velvetonu (opens in a new tab)