Geology topics

Carla Gomes

Publications and source records attributed to Carla Gomes.

4 recordsLinked to original sources

USGS2021

Learning augmented methods for matching: Improving invasive species management and urban mobility

With the success of machine learning, integrating learned models into real-world systems has become a critical chal- lenge. Naively applying predictions to combinatorial opti- mization problems can incur high costs, which has motivated researchers to consider learning augmented algorithms that can make use of faulty or incomplete predictions. Inspired by two matching problems in computational sustainability where data are abundant, we consider the learning augmented min weight matching problem where some nodes are revealed online while others are known a priori, e.g., by being pre- dicted by machine learning. We develop an algorithm that is able to make use of this extra information and provably im- proves upon pessimistic online algorithms. We evaluate our algorithm on two settings from computational sustainability – the coordination of opportunistic citizen scientists for inva- sive species management and the matching between taxis and riders under uncertain trip duration predictions. In both cases, we perform extensive experiments on real-world datasets and find that our method outperforms baselines, showing how learning augmented algorithms can reliably improve solu- tions for problems in computational sustainability

New York↗

USGS2021

Accelerating ecological sciences from above: Spatial contrastive learning for remote sensing

The rise of neural networks has opened the door for automatic analysis of remote sensing data. A challenge to using this machinery for computational sustainability is the necessity of massive labeled data sets, which can be cost-prohibitive for many non-profit organizations. The primary motivation for this work is one such problem; the efficient management of invasive species -- invading flora and fauna that are estimated to cause damages in the billions of dollars annually. As an ongoing collaboration with the New York Natural Heritage Program, we consider the use of unsupervised deep learning techniques for dimensionality reduction of remote sensing images, which can reduce sample complexity for downstream tasks and decreases the need for large labeled data sets. We consider spatially augmenting contrastive learning by training neural networks to correctly classify two nearby patches of a landscape as such. We demonstrate that this approach improves upon previous methods and naive classification for a large-scale data set of remote sensing images derived from invasive species observations obtained over 30 years. Additionally, we simulate deployment in the field via active learning and evaluate this method on another important challenge in computational sustainability -- landcover classification -- and again find that it outperforms previous baselines.

Proceedings of the AAAI Conference on Artificial I↗

USGS2019

Computational sustainability: Computing for a better world and a sustainable future

Computational sustainability aims to develop computational methods to help solve environmental, economic, and societal problems and thereby facilitate a path towards a sustainable future. Sustainability problems are unique in scale, impact, complexity, and richness, offering challenges but also opportunities for the advancement of the state of the art of computing and information science.

Communications of the ACM↗

USGS2019

Reserve design to optimize functional connectivity and animal density

Ecological distance-based spatial capture–recapture models (SCR) are a promising approach for simultaneously estimating animal density and connectivity, both of which affect spatial population processes and ultimately species persistence. We explored how SCR models can be integrated into reserve-design frameworks that explicitly acknowledge both the spatial distribution of individuals and their space use resulting from landscape structure. We formulated the design of wildlife reserves as a budget-constrained optimization problem and conducted a simulation to explore 3 different SCR-informed optimization objectives that prioritized different conservation goals by maximizing the number of protected individuals, reserve connectivity, and density-weighted connectivity. We also studied the effect on our 3 objectives of enforcing that the space-use requirements of individuals be met by the reserve for individuals to be considered conserved (referred to as home-range constraints). Maximizing local population density resulted in fragmented reserves that would likely not aid long-term population persistence, and maximizing the connectivity objective yielded reserves that protected the fewest individuals. However, maximizing density-weighted connectivity or preemptively imposing home-range constraints on reserve design yielded reserves of largely spatially compact sets of parcels covering high-density areas in the landscape with high functional connectivity between them. Our results quantify the extent to which reserve design is constrained by individual home-range requirements and highlight that accounting for individual space use in the objective and constraints can help in the design of reserves that balance abundance and connectivity in a biologically relevant manner.

Conservation Biology↗