arXiv subjects

Emma Collins

Publications and source records attributed to Emma Collins.

3 recordsLinked to original sources

arXiv2025

Geometrically Regularized Transfer Learning with On-Manifold and Off-Manifold Perturbation

Transfer learning under domain shift remains a fundamental challenge due to the divergence between source and target data manifolds. In this paper, we propose MAADA (Manifold-Aware Adversarial Data Augmentation), a novel framework that decomposes adversarial perturbations into on-manifold and off-manifold components to simultaneously capture semantic variation and model brittleness. We theoretically demonstrate that enforcing on-manifold consistency reduces hypothesis complexity and improves generalization, while off-manifold regularization smooths decision boundaries in low-density regions. Moreover, we introduce a geometry-aware alignment loss that minimizes geodesic discrepancy between source and target manifolds. Experiments on DomainNet, VisDA, and Office-Home show that MAADA consistently outperforms existing adversarial and adaptation methods in both unsupervised and few-shot settings, demonstrating superior structural robustness and cross-domain generalization.

cs.CV↗

arXiv2025

Disentangled Geometric Alignment with Adaptive Contrastive Perturbation for Reliable Domain Transfer

Despite progress in geometry-aware domain adaptation, current methods such as GAMA still suffer from two unresolved issues: (1) insufficient disentanglement of task-relevant and task-irrelevant manifold dimensions, and (2) rigid perturbation schemes that ignore per-class alignment asymmetries. To address this, we propose GAMA++, a novel framework that introduces (i) latent space disentanglement to isolate label-consistent manifold directions from nuisance factors, and (ii) an adaptive contrastive perturbation strategy that tailors both on- and off-manifold exploration to class-specific manifold curvature and alignment discrepancy. We further propose a cross-domain contrastive consistency loss that encourages local semantic clusters to align while preserving intra-domain diversity. Our method achieves state-of-the-art results on DomainNet, Office-Home, and VisDA benchmarks under both standard and few-shot settings, with notable improvements in class-level alignment fidelity and boundary robustness. GAMA++ sets a new standard for semantic geometry alignment in transfer learning.

cs.CV↗

arXiv2019

Mapping malaria seasonality: a case study from Madagascar

Many malaria-endemic areas experience seasonal fluctuations in case incidence as Anopheles mosquito and Plasmodium parasite life cycles respond to changing environmental conditions. While most existing maps of malaria seasonality use fixed thresholds of rainfall, temperature, and/or vegetation indices to identify suitable transmission months, we develop a statistical modelling framework for characterising the seasonal patterns derived directly from case data. The procedure involves a spatiotemporal regression model for estimating the monthly proportions of total annual cases and an algorithm to identify operationally relevant characteristics such as the transmission start and peak months. A seasonality index combines the monthly proportion estimates and existing estimates of annual case incidence to provide a summary of "how seasonal" locations are relative to their surroundings. An advancement upon past seasonality mapping endeavours is the presentation of the uncertainty associated with each map, which will enable policymakers to make more statistically sound decisions. The methodology is illustrated using health facility data from Madagascar.

stat.AP↗