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 “"Selective classification"”.
-
Two studies in resource-efficient inference: structural testing of networks, and selective classification
… to recovery. The second theme concerns using Selective classification (SC), or classification with an option to abstain, to control inference-time costs in the machine learning framework. The proposal is to learn a low-complexity selective classifier that only abstains on hard instances, and …
-
Robustness and safety of deep learning models.
… image watermarks. Then, we discuss the idea of selective classification (prediction with a reject option) to accept the imperfection of DL models and make best use of them. We propose a confidence score using raw logit output from the DL classifiers and show its better potential in performing …
-
Deep Learning-Based Medical Image Analysis Methods for Reducing Annotation Costs and Predictive Triage Misdiagnosis
… with various query strategies for nuclei classification using convolutional neural networks (CNN) trained on small patches containing single nuclei.<br/> <br/>This thesis proposes a multi-directional modification to the contrastive predictive coding (CPC) method for unsupervised …
-
Identifying regions of trusted prediction
… of prior work on conformal predictions and selective classification, we provide a unifying framework for confidence requirements that allows for distinguishing between various sources of uncertainty in the learning process as well as various region specifications. We then consider a set of …
-
Towards Reliable Hybrid Human-Machine Classifiers
… classifiers to be deployed in cost-sensitive classification tasks. The objective is to assess ML quality in hybrid classification contexts and design the appropriate metrics, thereby knowing whether we can trust the model predictions and identifying the subset of items on which the model is …
-
Uncertainty estimation for adaptable and reliable robotic vision
… is shown to perform well in the problem of selective classification in the presence of out-of-distribution data.
-
Uncertainty-Driven Reliability: Selective Prediction and Trustworthy Deployment in Modern Machine Learning
… of ML, with a particular focus on selective prediction—a paradigm where models abstain from predicting when confidence is low. We first show that a model's training trajectory contains rich uncertainty signals that can be exploited without altering its architecture or loss. By …
-
From Deterministic to Uncertainty-Aware Neural Networks: A From-Scratch Study on MNIST and Fashion-MNIST
… προβλεπτική επίδοση. Πρόσθετες αναλύσεις, όπως η selective classification, το temperature scaling και η αξιολόγηση υπο συνθήκες εκτός κατανομής, δείχνουν ότι οι εκτιμήσεις αβεβαιότητας παρέχουν χρήσιμη πληροφορία για την αξιοπιστία των προ- βλέψεων. Το βασικό συμπέρασμα είναι ότι ο λόγος που οι …
-
Identifying Age and Gender and Analysing their Relation to Programming Behaviour of Scratch Users
… and a text-based Transformer model with selective classification for age identification. Cross-validation results revealed that both of these models achieve an F1-score of around 0.80 on the training set. We used these models to automatically elicit age and gender information for the rest …
-
On learning, fairness, and complexity
… fairness framework and the problem of learning selective classifiers, which are predictors that are allowed to abstain on some fraction of the domain. Building on the notion of omniprediction (which is in turn built using tools from the multigroup fairness framework), given a pre-specified class …
-
Novel neural architectures & algorithms for efficient inference
… to others. Input hardness enables us to create selective classifiers wherein a low-capacity network handles simple inputs while abstaining from a prediction on the complex inputs. Next, we create hybrid models that route the hard inputs from the low-capacity abstaining network to a high-capacity …