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 20 of 71 for “"Influence Functions"”.
-
Influence functions in multivariate analysis
In this thesis we derive and apply influence functions for the detection of observations in multivariate analysis which when omitted from, or added to, the data lead to substantial changes in some aspect of our analysis. Emphasis is placed on the influence functions for the eigenvalues and …
-
Influence functions - comparison of existing methods /
… of them is a technique from robust statistics - influence functions (IF). However, many papers indicate that IF return outliers and mislabelled data as the most influential points from training set. In order to identify relevant examples that are more understandable for end users, alternative …
-
Influence functions, higher moments, and hedging
This thesis includes three chapters regarding influence functions, higher moments, and futures hedging. In Chapter 2, the objective is to use an influence function to better understand semi-kurtosis for use in analyzing peakedness and tail heaviness on one side of a distribution. Also, it is shown …
-
Improving Explainability and Interpretability of Neural Networks via Hyperparameter-Extended Influence Functions
… for explainability by generalizing classical influence functions to encompass user-defined hyperparameters embedded in the training loss, model architecture, or data representation. By extending influence functions in this way, the framework broadens their applicability and integrates multiple …
-
Weight Calibration in the Joint Modelling of Medical Cost and Survival for Complex Samples
… We apply survey calibration to surrogates of the influence functions as calibration variables in a joint model setting, which involves multiple sets of surrogates of the influence functions. We show that calibration to surrogates of the survival model influence functions leads to efficiency gains …
-
Influence-Function-Based Causal Inference and Sensitivity Analysis
… inference and diagnostic tools grounded in influence-function theory. Originating in robust statistics, influence functions measure local sensitivity of estimators to changes in the underlying data-generating process, enabling researchers to evaluate efficiency, construct valid confidence …
-
Beyond Gradients: Using Curvature Information for Deep Learning
… focusing on hyperparameter optimization and influence functions. Bilevel optimization involves nested optimization problems where the outer-level objective depends on the optimal solution to an inner-level optimization. In gradient-based bilevel optimization, a key challenge is accurately …
-
BETTER MODELS FOR HIGH-STAKES TASKS
… to deliver instance-level explanations using influence functions. Next, we show that these influence functions function are fairly robust across domains. Then, we develop an efficient method that reduces model uncertainty while modeling data uncertainty via Bayesian Neural Networks. Finally, …
-
A Unified Non-Visual Explainable AI Framework for Neonatal Pain Assessment
… diagnosis for neonatal pain assessment using influence functions applied to neonatal audio spectrogram images. We employ influence functions to quantify the impact of individual training samples on model predictions. Experimental results demonstrate that the proposed approach can detect and …
-
Using finite mixtures to robustify statistical models
… of the proposed estimates using the concept of influence function. The estimates are defined iteratively, so that the implicit differentiation method of Jorgensen is used to obtain the influence functions of the estimates. We give example plots of these influence functions which are bounded. In …
-
Semi-parametric Estimation in the Block Parallel Missing Data Graphical Model
… (MNAR) model, in particular estimator based on Influence Functions (IFs) in the Block Parallel Missing Data graphical model. Experiments show that estimators based on Influence Function (IF) make a significant improvement over more common Inverse-Probability Weighted (IPW) estimators, especially …
-
Evaluating and Mitigating Implicit Bias in Large Language Models
… during the fine-tuning stage. The work leverages influence functions to score each training sample on how influential it is on the model's biased outputs. Although influence functions have been there in the community for long, this work explores a novel way to make them computationally efficient …
-
Vedybų modeliavimas: skyrybų nuspėjimas ir vedybų problemų sprendimas /
… the problems that arise in marriages. We analyse influence functions, as well as deterministic and stochastic models to understand the complex dynamics of relationships. Mathematical models explore features including emotional inertia, equilibrium points, external and partner influences to reflect …
-
The order of things : short stories
… The distance between narrator and character in "Influence" functions as a way of understanding the distance the characters experience between each other and from themselves. The second story, "The Order of Things", explores similar themes of alienation and miscommunication through a more intimate …
-
STRESS FIELDS NEAR HOLES OR CUT-OUTS IN A UNIDIRECTIONAL FIBER REINFORCED COMPOSITE LAMINA
… fiber break solutions, we apply the break influence superposition technique (BIS), developed in the context of the shear-lag model by Sastry, Beyerlein and Phoenix (1996), to calculate the stress field in the composite. Stress fields for three different hole shapes have been studied and …
-
Improvement in Probabilistic Information Retrieval Model: Rewarding Terms with High Relative Term Frequency
… in other words, I introduce a set of three influence functions with the application of relative term frequency to model and enhance the performance of the fundamental probabilistic weighting function, BM25. The study aims to exploit the properties of the combination of relative term …
-
Principled and Explainable Reasoning involving Natural Languages
… of reasoning systems, we study how influence functions can be used to perform training data attribution for text generation models. There are two challenges in this task -- first it is unclear how to use influence functions on the sequential output; second estimation of influence …
-
On Theory of Joint Influence
Joint Influence is a kind of phenomena that can be commonly observed in many areas of application. For example, in physics, a particle receives forces from a number of other particles, and such forces jointly effect the motion of the particle. This phenomenon also arises in areas like social …
Page 1 of 4