Sökning: "latent space exploration"

Visar resultat 1 - 5 av 6 uppsatser innehållade orden latent space exploration.

  1. 1. Topological regularization and relative latent representations

    Master-uppsats, KTH/Skolan för elektroteknik och datavetenskap (EECS)

    Författare :Alejandro García Castellanos; [2023]
    Nyckelord :Algebraic Topology; Large Language Models; Relative Representation; Representation Learning; Model Stitching; Topological DataAnalysis; Zero-shot; Algebraisk topologi; Stora språkmodeller; Relativ representation; Representationsinlärning; Modell sömmar; Topologisk dataanalys; Zero-shot;

    Sammanfattning : This Master's Thesis delves into the application of topological regularization techniques and relative latent representations within the realm of zero-shot model stitching. Building upon the prior work of Moschella et al. LÄS MER

  3. 2. Improving Change Point Detection Using Self-Supervised VAEs : A Study on Distance Metrics and Hyperparameters in Time Series Analysis

    Master-uppsats, KTH/Skolan för elektroteknik och datavetenskap (EECS)

    Författare :Daniel Workinn; [2023]
    Nyckelord :Change point detection; Time series data; Segmentation; Machine learning; Data mining; Detektion av brytpunkter; Tidsseriedata; Segmentering; Maskininlärning; Datautvinning;

    Sammanfattning : This thesis addresses the optimization of the Variational Autoencoder-based Change Point Detection (VAE-CP) approach in time series analysis, a vital component in data-driven decision making. We evaluate the impact of various distance metrics and hyperparameters on the model’s performance using a systematic exploration and robustness testing on diverse real-world datasets. LÄS MER

  4. 3. Messing With The Gap: On The Modality Gap Phenomenon In Multimodal Contrastive Representation Learning

    Uppsats för yrkesexamina på avancerad nivå, Uppsala universitet/Industriell teknik

    Författare :Mohammad Al-Jaff; [2023]
    Nyckelord :Multimodal machine learning; Representation learning; Self-supervised learning; contrastive learning; computer vision; computational biology; bioinformatics;

    Sammanfattning : In machine learning, a sub-field of computer science, a two-tower architecture model is a specialised type of neural network model that encodes paired data from different modalities (like text and images, sound and video, or proteomics and gene expression profiles) into a shared latent representation space. However, when training these models using a specific contrastive loss function, known as the multimodalinfoNCE loss, seems to often lead to a unique geometric phenomenon known as the modality gap. LÄS MER

  5. 4. An Application of LatentCF++ on Providing Counterfactual Explanations for Fraud Detection

    Master-uppsats, Stockholms universitet/Institutionen för data- och systemvetenskap

    Författare :Maria-Sofia Giannopoulou; [2023]
    Nyckelord :counterfactuals; fraud detection; LatentCF ; autoencoder; one-dimensional convolutional neural network;

    Sammanfattning : The aim of explainable machine learning is to aid humans in understanding how exactly complex machine learning models work. Machine learning models have offered great performance in various areas. However, the mechanisms behind how the model works and how decisions are being made remain unknown. LÄS MER

  7. 5. Basil-GAN

    Master-uppsats, KTH/Matematisk statistik

    Författare :Jonatan Risberg; [2022]
    Nyckelord :GAN; mathematical statistics; deep neural networks; generative models; latent space exploration; sequential data; GAN; matematisk statistik; djupa neurala nätverk; generativa modeller; utforskning av latenta rum; sekventiell data;

    Sammanfattning : Developments in computer vision has sought to design deep neural networks which trained on a large set of images are able to generate high quality artificial images which share semantic qualities with the original image set. A pivotal shift was made with the introduction of the generative adversarial network (GAN) by Goodfellow et al.. LÄS MER