홍석쓰 블로그 https://redstarhong.tistory.com/ Hongsuk Choi ko Wed, 6 May 2026 17:57:56 +0900 TISTORY 100 홍돌 홍석쓰 블로그 https://tistory1.daumcdn.net/tistory/2219311/attach/46676e92b3a84a128b820c8b94b5e6d6 https://redstarhong.tistory.com Retargeting and Optimization https://redstarhong.tistory.com/359 And solving the linear equation system when the system matrix is a triangular matrix is very efficient! # Normal equations: (J^TJ)x = J^Ty # This is exactly our Ax = b where: A = J^TJ # symmetric positive definite! b = J^Ty # Solve using Cholesky: A = .. Research (연구 관련) 홍돌 https://redstarhong.tistory.com/359 https://redstarhong.tistory.com/359#entry359comment Mon, 17 Feb 2025 12:06:35 +0900 Inference-Time Techniques for LLM Reasoning https://redstarhong.tistory.com/358 Lecture 1, Jan 27th, 2025, https://rdi.berkeley.edu/adv-llm-agents/slides/inference_time_techniques_lecture_sp25.pdf   Background: How do we evaluate consistency of free-form answers from multiple answer generation? Research (연구 관련) 홍돌 https://redstarhong.tistory.com/358 https://redstarhong.tistory.com/358#entry358comment Mon, 3 Feb 2025 08:24:24 +0900 Dual Contouring https://redstarhong.tistory.com/357 What is Dual contouring? Dual contouring (DC) is a popular isosurface extraction (surface reconstruction) algorithm for converting a volumetric representation (e.g., an implicit field or voxel grid) into a polygon mesh. It is called “dual” bec.. Research (연구 관련) 홍돌 https://redstarhong.tistory.com/357 https://redstarhong.tistory.com/357#entry357comment Fri, 17 Jan 2025 18:36:48 +0900 Generative AI - Diffusion / Lecture 1 https://redstarhong.tistory.com/356 Jitendra lunch group meeting; source: (Israel / The Technion) Probability Theory 101 What is the "convolution property of a PDF (probability density function)"? For two independent random variables, the PDF of their sum is the convolution o.. Research (연구 관련) 홍돌 https://redstarhong.tistory.com/356 https://redstarhong.tistory.com/356#entry356comment Wed, 27 Nov 2024 09:48:12 +0900 Transformer / Large models https://redstarhong.tistory.com/355 What is KV caching?  KV caching is specifically related to the auto-regressive approach of a transformer decoder. In a transformer decoder, it attends to the past and current tokens, but not to future tokens. At each time step, the transformer repeat.. Research (연구 관련) 홍돌 https://redstarhong.tistory.com/355 https://redstarhong.tistory.com/355#entry355comment Mon, 25 Nov 2024 14:47:14 +0900