MLPR 연구실에서는 Machine Learning과 Pattern Recognition을 활용해서 효율적인 인공지능 알고리즘 또는 모델을 만드는 연구를 하고 있습니다. 특히, Ubiqutous AI "누구에게나 어디에서든지 사용가능한 AI"를 지향하고 있습니다.
We are interested in Ubiqutous AI but not limited to:
AI in resource-constrained environments: meta-learning, domain generalization, domain adaptation
Model compression (on device learning): model pruning, model quantization, ensemble method and knowledge distillation
Application domain: generative model, natural language processing (NLP) and its downstream tasks, object segmentation
Meta-learning in resouce-constrained enviroments
Knowlede distillation
Model compression
Application: Object generation
The image on the left is the source image with a box given by hand and the image on the right is the result of our method which generates a vehicle inside the box.
Application: Adversarial sample detection in natural language processing
