Byungkon Kang

Dr. Byungkon Kang


Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
22 November 2023 in Electronics
Verified
Journal Article
Open access
PDF file
AVAlfredo Valenzuela
Alfredo Valenzuela
JCJongseong Choi
Jongseong Choi
RORicardo Ortiz
Ricardo Ortiz
BKByungkon Kang
Byungkon Kang
MKMichael Kim
Michael Kim
TKTaewook Kang
Taewook Kang
Unmanned ground vehicles (UGVs) have emerged as a promising solution for reconnaissance missions, overcoming labor cost, frequency, and subjectivity issues associated with manual procedures. However, for dynamic environments such as construction sites, the constantly changing conditions hinder a manager from planning the UGV’s paths. For an autonomous monitoring mission, the path planning should be dealt with by having a map with the site’s most recent scene. In this study, we develop a mobile app capable of local map updates by overlaying an image on a mapping API (e.g., Google Maps) hence working as a digital twin capable of creating a dynamic representation of the updated terrain over the mapping API. UGV operators can draw a path on such an updated construction scene using a tablet PC or smartphone. Discrete GPS information (e.g., latitudinal, and longitudinal) is executed for the UGV’s controller. In the overlaying procedure, the homographic relation between the image and map is automatically computed and then projected. We successfully demonstrated the capabilities of the technique with two construction sites and a soccer field using images from the ground and satellite, respectively. The error generation is quantitatively measured and analyzed.


17 July 2024 in arXiv
Verified
Preprint
Open access
PDF file
HKHa Young Kim
Ha Young Kim
NBNiranjan Balasubramanian
Niranjan Balasubramanian
BKByungkon Kang
Byungkon Kang
It has become common practice now to use random initialization schemes, rather than the pre-trained embeddings, when training transformer based models from scratch. Indeed, we find that pre-trained word embeddings from GloVe, and some sub-word embeddings extracted from language models such as T5 and mT5 fare much worse compared to random initialization. This is counter-intuitive given the well-known representational and transfer-learning advantages of pre-training. Interestingly, we also find that BERT and mBERT embeddings fare better than random initialization, showing the advantages of pre-trained representations. In this work, we posit two potential factors that contribute to these mixed results: the model sensitivity to parameter distribution and the embedding interactions with position encodings. We observe that pre-trained GloVe, T5, and mT5 embeddings have a wider distribution of values. As argued in the initialization studies, such large value initializations can lead to poor training because of saturated outputs. Further, the larger embedding values can, in effect, absorb the smaller position encoding values when added together, thus losing position information. Standardizing the pre-trained embeddings to a narrow range (e.g. as prescribed by Xavier) leads to substantial gains for Glove, T5, and mT5 embeddings. On the other hand, BERT pre-trained embeddings, while larger, are still relatively closer to Xavier initialization range which may allow it to effectively transfer the pre-trained knowledge.