arXiv subjects

Holly Jackson

Publications and source records attributed to Holly Jackson.

4 recordsLinked to original sources

arXiv2026

Depth from Defocus via Direct Optimization

Though there exists a reasonable forward model for blur based on optical physics, recovering depth from a collection of defocused images remains a computationally challenging optimization problem. In this paper, we show that with contemporary optimization methods and reasonable computing resources, a global optimization approach to depth from defocus is feasible. Our approach rests on alternating minimization. When holding the depth map fixed, the forward model is linear with respect to the all-in-focus image. When holding the all-in-focus image fixed, the depth at each pixel can be computed independently, enabling embarrassingly parallel computation. We show that alternating between convex optimization and parallel grid search can effectively solve the depth-from-defocus problem at higher resolutions than current deep learning methods. We demonstrate our approach on benchmark datasets with synthetic and real defocus blur and show promising results compared to prior approaches. Our code is available at github.com/hollyjackson/dfd.

cs.CV↗

arXiv2022

Evolution Gym: A Large-Scale Benchmark for Evolving Soft Robots

Both the design and control of a robot play equally important roles in its task performance. However, while optimal control is well studied in the machine learning and robotics community, less attention is placed on finding the optimal robot design. This is mainly because co-optimizing design and control in robotics is characterized as a challenging problem, and more importantly, a comprehensive evaluation benchmark for co-optimization does not exist. In this paper, we propose Evolution Gym, the first large-scale benchmark for co-optimizing the design and control of soft robots. In our benchmark, each robot is composed of different types of voxels (e.g., soft, rigid, actuators), resulting in a modular and expressive robot design space. Our benchmark environments span a wide range of tasks, including locomotion on various types of terrains and manipulation. Furthermore, we develop several robot co-evolution algorithms by combining state-of-the-art design optimization methods and deep reinforcement learning techniques. Evaluating the algorithms on our benchmark platform, we observe robots exhibiting increasingly complex behaviors as evolution progresses, with the best evolved designs solving many of our proposed tasks. Additionally, even though robot designs are evolved autonomously from scratch without prior knowledge, they often grow to resemble existing natural creatures while outperforming hand-designed robots. Nevertheless, all tested algorithms fail to find robots that succeed in our hardest environments. This suggests that more advanced algorithms are required to explore the high-dimensional design space and evolve increasingly intelligent robots -- an area of research in which we hope Evolution Gym will accelerate progress. Our website with code, environments, documentation, and tutorials is available at http://evogym.csail.mit.edu.

cs.RO↗

arXiv2020

Using heritability of stellar chemistry to reveal the history of the Milky Way

Since chemical abundances are inherited between generations of stars, we use them to trace the evolutionary history of our Galaxy. We present a robust methodology for creating a phylogenetic tree, a biological tool used for centuries to study heritability. Combining our phylogeny with information on stellar ages and dynamical properties, we reconstruct the shared history of 78 stars in the Solar Neighbourhood. The branching pattern in our tree supports a scenario in which the thick disk is an ancestral population of the thin disk. The transition from thick to thin disk shows an anomaly, which we attribute to a star formation burst. Our tree shows a further signature of the variability in stars similar to the Sun, perhaps linked to a minor star formation enhancement creating our Solar System. In this paper, we demonstrate the immense potential of a phylogenetic perspective and interdisciplinary collaboration, where with borrowed techniques from biology we can study key processes that have contributed to the evolution of the Milky Way.

astro-ph.GA↗

arXiv2019

Traits for chemical evolution in solar twins: Trends of neutron-capture elements with stellar age

The physical processes driving chemical evolution in the Milky Way can be probed using the distribution of abundances in low-mass FGK type stars in space phase at different times. During their final stages of evolution stars experience nucleosynthesis several times, each at different timescales and producing different chemical elements, so finding abundance ratios that have simple variations across cosmic times remains a challenge. Using the sample of 80 solar twins for which ages and abundances of 30 elements were measured with high precision by Bedell et al 2018, we searched for all possible abundance ratios combinations that show linear trends with age. We found 55 such ratios, all combining a neutron-capture element and another one produced by different nucleosynthesis channels. We recovered the ratios of [Y/Mg], [Ba/Mg] and [Al/Y] reported previously in the literature, and found that [C/Ba] had the largest dependency with age with a slope of $0.049 \pm 0.003~\mathrm{dex/Gyr}$, imposing constraints on the magnitude of time dependency of abundance ratios in solar twins. Our results suggest using s-process elements, in lieu of Fe, as a reference for constraining chemical evolution models of the solar neighbourhood. Our study illustrates hows a large variety of chemical elements measured from high resolution spectra is key in facing current challenges in understanding the formation and evolution of our Galaxy.

astro-ph.SR↗