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 18 of 18 for “"sparse autoencoders"”.

  1. Interpreting multimodal speech-vision models with sparse autoencoders

    … modalities remains difficult. This work uses Sparse Autoencoders (SAEs) to interpret a self-supervised audio-visual model, focusing on spoken language rather than text as the linguistic modality. We train and compare four SAE architectures (Standard, JumpReLU, BatchTopK, and Matryoshka) on …

    ubc Repository record for Interpreting multimodal speech-vision models with sparse autoencoders (opens in a new tab) Download PDF for Interpreting multimodal speech-vision models with sparse autoencoders (opens in a new tab)

  2. Sparse Autoencoder Enhanced Universal Image Restoration with DACLIP and Diffusion Models

    … This thesis presents SAE-Restore, a dual-stage sparse autoencoder framework addressing these limitations through explicit feature denoising and image refinement. SAE-Restore integrates two sparse autoencoders at strategic positions. SAE1 performs feature-space denoising after cross-attention …

    unt Repository record for Sparse Autoencoder Enhanced Universal Image Restoration with DACLIP and Diffusion Models (opens in a new tab)

  3. Diffusion models as representation learners: interpreting and leveraging their latent features

    … we present Revelio, a framework that uses k-Sparse Autoencoders (k-SAE) to extract monosemantic, human-interpretable features and analyze how semantic concepts emerge across layers, timesteps, and architectures. We complement this with a discriminative analysis using lightweight classifiers …

    bu Repository record for Diffusion models as representation learners: interpreting and leveraging their latent features (opens in a new tab)

  4. Recklessly approximate sparse coding

    … nets, factored RBMs, mcRBMs, convolutional RBMs, sparse autoencoders and several others. Moreover, these features are extremely simple and easy to compute. Several intuitive arguments have been put forward to describe this remarkable performance, yet no mathematical justification has been offered. …

    ubc Repository record for Recklessly approximate sparse coding (opens in a new tab) Download PDF for Recklessly approximate sparse coding (opens in a new tab)

  5. Unsupervised Representation Learning with Autoencoders

    … of learning unsupervised representations using autoencoders, and propose regularization techniques that enable autoencoders to learn useful representations of data in unsupervised and semi-supervised settings. First, we exploit sparsity as a generic prior on the representations of autoencoders …

    toronto-retro Repository record for Unsupervised Representation Learning with Autoencoders (opens in a new tab)

  6. Interpretable computational metaphor processing

    … models (LLMs) for metaphor understanding through sparse autoencoders and dictionary learning, decomposing latent representations to extract monosemantic features that improve metaphor transparency. The findings contribute to the development of more linguistically sophisticated, contextually …

    whiterose Download PDF for Interpretable computational metaphor processing (opens in a new tab)

  7. Interpretation and Control of AI Model Behavior Through Direct Adjustment of Latent Representations

    … token positions. Similar experiments using sparse autoencoders in place of linear probes yielded significantly poorer performance. These results highlight the effectiveness of simple linear probes as valuable tools for interpretability and control.

    ucf Repository record for Interpretation and Control of AI Model Behavior Through Direct Adjustment of Latent Representations (opens in a new tab)

  8. Linear recurrent models for robust and interpretable long-context modeling

    … memory to bridge local attention windows. Using sparse autoencoders, we extract interpretable features from these recurrent states to explore their role in retaining long-range information. Through tasks such as long-context passkey retrieval, we reveal how these mechanisms facilitate extended …

    cornell Repository record for Linear recurrent models for robust and interpretable long-context modeling (opens in a new tab)

  9. Deep Learning Techniques for Image Segmentation of Whole Body MRI

    … train convolutional neural networks and stacked sparse autoencoders. The success of these deep learning models relies on the high contrast provided by multiparametric MRIs for distinguishing voxels of different classes. The models achieve excellent segmentation accuracy in slices of the thighs. …

    jhu Repository record for Deep Learning Techniques for Image Segmentation of Whole Body MRI (opens in a new tab)

  10. Efficient and Interpretable Representations: From Medical Representation Learning to Vision-Language Multimodal Representation Engineering

    … analysis of pathology foundation models using sparse autoencoders reveals a fundamental scalability problem: each model produces incompatible latent spaces that require separate expert analysis, creating exponential scaling in interpretability effort as foundation models proliferate. To address …

    concordia Download PDF for Efficient and Interpretable Representations: From Medical Representation Learning to Vision-Language Multimodal Representation Engineering (opens in a new tab)

  11. Geometric Deep Learned Descriptors for 3D Shape Recognition

    … retrieval using geodesic moments and stacked sparse autoencoders. The key idea is to learn deep shape representations in an unsupervised manner. Such discriminative shape descriptors can then be used to compute pairwise dissimilarities between shapes in a dataset, and to find the retrieved set …

    concordia Download PDF for Geometric Deep Learned Descriptors for 3D Shape Recognition (opens in a new tab)

  12. Modelos computacionales de movimiento ocular

    … Alzheimer, utilizando deep-learning con denoise-sparse-autoencoders y un clasifi cador, a partir de información del movimiento ocular durante la lectura. Los resultados obtenidos, con un 89;8% de efectividad en la clasi ficación por oración y 100% por sujeto, son satisfactorios. Esto sugiere que …

    uns-ar Repository record for Modelos computacionales de movimiento ocular (opens in a new tab)

  13. Integrated Graph Theoretic, Radiomics, and Deep Learning Framework for Personalized Clinical Diagnosis, Prognosis, and Treatment Response Assessment of Body Tumors

    … MPDL network was constructed from stacked sparse autoencoders (SSAE) with five hidden layers. The MPDL network parameters were optimized using k-fold cross-validation. In addition, the MPDL segmentation network was tested on an independent dataset. Feature Extraction: I developed the …

    jhu Repository record for Integrated Graph Theoretic, Radiomics, and Deep Learning Framework for Personalized Clinical Diagnosis, Prognosis, and Treatment Response Assessment of Body Tumors (opens in a new tab)

  14. Robust and Interpretable Medical Image Segmentation via Retrieval and Mechanistic Analysis

    … their intermediate representations using Sparse Autoencoders (SAEs). This decomposition yields sparse, disentangled latent features corresponding to recurring anatomical and pathological activation patterns, enabling direct inspection and manipulation of individual components. To bridge …

    wilfrid-laurier Repository record for Robust and Interpretable Medical Image Segmentation via Retrieval and Mechanistic Analysis (opens in a new tab) Download PDF for Robust and Interpretable Medical Image Segmentation via Retrieval and Mechanistic Analysis (opens in a new tab)

  15. Development of synthetic power distribution networks and datasets with industrial validation

    … distribution networks. Two new methods based on sparse autoencoders (SAEs), and multi-stage transfer learning (MSTL) are proposed for load profile classification. Different from conventional hand-crafted feature representations, SAEs can learn useful features from vast amounts of building data in …

    unsw Repository record for Development of synthetic power distribution networks and datasets with industrial validation (opens in a new tab)

  16. Suicide Ideation Detection from Social Media Using Language Models: Data Augmentation and Interpretability

    … representations learned by these models. Using sparse autoencoders and geometric analyses, including UMAP projections and cosine-distance measurements, we analyze whether psychologically meaningful risk factors are encoded as more distinct and separable directions in the models' latent spaces. …

    ottawa-retro Repository record for Suicide Ideation Detection from Social Media Using Language Models: Data Augmentation and Interpretability (opens in a new tab)

  17. Domain decomposition-based reduced-order models using nonlinear-manifolds and interpolatory projections

    This thesis integrates nonlinear-manifold and interpolatory projection model reduction with domain decomposition (DD) to reduce the offline costs of training reduced-order models (ROMs). In the neural network-based nonlinear-manifold ROM (NM-ROM) setting, the number of parameters requiring training …

    rice Repository record for Domain decomposition-based reduced-order models using nonlinear-manifolds and interpolatory projections (opens in a new tab)

  18. Příznaky z videa pro klasifikaci

    Tato práce porovnává ručně-navrženy příznaky s příznaky naučenými metodami učení příznaků při klasifikací videa. Příznaky naučené pomocí Analýzy nezávislých podprostorů, Řídkými Autoenkodéry a vybělením Analýzou hlavních komponent byly otestovány v systému pro klasifikaci videa pomocí Bag of Words, …

    brno-tech Repository record for Příznaky z videa pro klasifikaci (opens in a new tab)