arXiv Science⌕ Search

arXiv subjects

Jim Thomas

Publications and source records attributed to Jim Thomas.

4 recordsLinked to original sources

Balanced Twin Weight Factorization for Efficient MLLMs

Weight Factorization (WF) methods aim to reduce a model's latency, memory, or both. These typically decompose a pretrained model into two or more factors that approximate the original model weights with fewer parameters and floating point operations (FLOPs). However, translating fewer FLOPs into reduced latency is challenging for WF methods in practice due to overhead introduced by replacing a single large weight matrix with multiple smaller ones. To address this, we propose Balanced Twin Weight Factorization (BTWF), which uses the unique structure of factorized models to simultaneously lower latency and increase learning capacity. Specifically, BTWF operates in the shared basis WF setting, where two or more layers are decomposed into a shared basis matrix and a layer-specific mixing matrix. If these shared layers have the same input (i.e., are twins), then we need only perform this shared forward pass once, reducing inference time. However, twin layers are often of varying sizes, and methods from prior work tend to favor the larger layer. In contrast, our approach boosts performance by combining Factor Component Balancing (FCB), which ensures every shared member is well represented in the selected factors, with Block Contribution Balancing (BCB), which considers downstream model effects. Further, we also show that an overparameterized BTWF, where some layer's factors result in higher FLOPs than the original during training, can boost task performance without additional resources at test time. Experiments on six models over nine benchmarks show BTWF obtains a 2-6% gain over the state-of-the-art without expensive finetuning (1-3.5% gain with it) at comparable computational costs.

cs.CV↗

Learning Feature Decomposition for Domain Adaptive Monocular Depth Estimation

Monocular depth estimation (MDE) has attracted intense study due to its low cost and critical functions for robotic tasks such as localization, mapping and obstacle detection. Supervised approaches have led to great success with the advance of deep learning, but they rely on large quantities of ground-truth depth annotations that are expensive to acquire. Unsupervised domain adaptation (UDA) transfers knowledge from labeled source data to unlabeled target data, so as to relax the constraint of supervised learning. However, existing UDA approaches may not completely align the domain gap across different datasets because of the domain shift problem. We believe better domain alignment can be achieved via well-designed feature decomposition. In this paper, we propose a novel UDA method for MDE, referred to as Learning Feature Decomposition for Adaptation (LFDA), which learns to decompose the feature space into content and style components. LFDA only attempts to align the content component since it has a smaller domain gap. Meanwhile, it excludes the style component which is specific to the source domain from training the primary task. Furthermore, LFDA uses separate feature distribution estimations to further bridge the domain gap. Extensive experiments on three domain adaptative MDE scenarios show that the proposed method achieves superior accuracy and lower computational cost compared to the state-of-the-art approaches.

cs.CV↗

Design and implementation of wire tension measurement system for MWPCs used in the STAR iTPC upgrade

The STAR experiment at RHIC is planning to upgrade the Time Projection Chamber which lies at the heart of the detector. We have designed an instrument to measure the tension of the wires in the multi-wire proportional chambers (MWPCs) which will be used in the TPC upgrade. The wire tension measurement system causes the wires to vibrate and then it measures the fundamental frequency of the oscillation via a laser based optical platform. The platform can scan the entire wire plane, automatically, in a single run and obtain the wire tension on each wire with high precision. In this paper, the details about the measurement method and the system setup will be described. In addition, the test results for a prototype MWPC to be used in the STAR-iTPC upgrade will be presented.

physics.ins-det↗

Chiral Magnetic Effect Task Force Report

In this report, we briefly examine the current status of the study of the chiral magnetic effect including theory and experimental progress. We recommend future strategies for resolving uncertainties in interpretation including recommendations for theoretical work, recommendations for measurements based on data collected in the past five years, and recommendations for beam use in the coming years of RHIC. We have specifically investigated the case for colliding nuclear isobars (nuclei with the same mass but different charge) and find the case compelling. We recommend that a program of nuclear isobar collisions to isolate the chiral magnetic effect from background sources be placed as a high priority item in the strategy for completing the RHIC mission.

nucl-th↗