A5074567888- Random lasers and scattering media
- Astro and Planetary Science
- Aerospace Engineering and Energy Systems
- Digital Holography and Microscopy
- Spacecraft Dynamics and Control
- Topic Modeling
- Natural Language Processing Techniques
- Optical Coherence Tomography Applications
- Speech Recognition and Synthesis
- Advanced Optical Sensing Technologies
- Advanced X-ray Imaging Techniques
University of California, Berkeley
2019-2020
Mask-based lensless imagers are smaller and lighter than traditional lensed cameras. In these imagers, the sensor does not directly record an image of scene; rather, a computational algorithm reconstructs it. Typically, mask-based use model-based reconstruction approach that suffers from long compute times heavy reliance on both system calibration heuristically chosen denoisers. this work, we address limitations using bounded-compute, trainable neural network to reconstruct image. We...
Titan's dense atmosphere, low gravity, and high winds at altitudes create descent times of >90 minutes with standard entry/descent/landing (EDL) architectures result in large unguided landing ellipses, 99% values ~110×110 km 149×72 recent Titan lander proposals. Enabling precision on could increase science return for the types missions proposed to date make additional sites accessible, opening up new possibilities investigations. Precision has unique challenges, because hazy atmosphere makes...
Parameter-efficient methods are able to use a single frozen pre-trained large language model (LLM) perform many tasks by learning task-specific soft prompts that modulate behavior when concatenated the input text. However, these learned tightly coupled given -- if is updated, corresponding new need be obtained. In this work, we propose and investigate several approaches "Prompt Recycling'" where prompt trained on source transformed work with target model. Our do not rely supervised pairs of...