About — Chi-Wei Lee

L2/3

In my own words關於我

I graduated from National Tsing Hua University with a double major in Physics and Electrical Engineering & Computer Science (AI Track).

My theory work connects predictive coding, associative memory, and Bayesian inference. Two co-first-author manuscripts are under review: one identifies associative memory as the denoiser predictive coding requires; the other replaces point-estimate inference with preconditioned Langevin sampling.

At NTHU's Human-Centered Machine Intelligence Lab, I combine computer vision and generative modelling for 3D reconstruction from fMRI—decoding the spatial world represented by neural activity while quantifying uncertainty. In parallel, I have worked on controllable diffusion at Academia Sinica and robustness in chest X-ray classification. I am now one of nine delegates selected nationally for a 60-day AI × Brain and Neuroscience research placement at UCLA through Taiwan's Global Pathfinders Initiative, and I am looking for NeuroAI PhD positions for 2027 entry.

我畢業於國立清華大學，雙主修物理系與電機資訊學院學士班（AI 組）。

我的理論研究連結預測編碼、聯想記憶與貝氏推論。兩篇共同第一作者手稿正在審查：一篇指出聯想記憶正是預測編碼所需的去噪器；另一篇以 preconditioned Langevin sampling 取代只回傳單一點估計的推論。

在清華 HMI Lab，我結合電腦視覺與生成模型，進行從 fMRI 到 3D 空間的重建：解碼神經活動所表徵的空間世界，同時量化重建的不確定性。我也曾在中研院研究可控 diffusion，並投入胸腔 X 光分類的穩健性問題。目前透過教育部青年百億海外圓夢基金計畫，以全國九位代表之一的身分赴 UCLA 進行 60 天 AI × Brain and Neuroscience 研究，並尋找 2027 年入學的 NeuroAI 博士班機會。