IEEE

A Bayesian Approach to OOD Robustness in Image Classification

Abstract: An important and unsolved problem in computer vision is to ensure that the algorithms are robust to changes in image domains. We address this problem in the scenario where we only have ...
IEEE

A Bayesian Backed DeepLabV3+ Customized Hybrid-Depth Framework for Brain Tumor Segmentation and Classification

Abstract: Efficient diagnosis, treatment planning, and patient care requires that brain tumors in MRI images must be accurately segmented and classified. Many deep learning practices are used to ...
note

The First-Year Hurdles of a Resource Room Teacher (Part 4): The Day I Froze While Writing in the Communication Notebook — Three Lines to Connect Hearts, Rather Than One ...

There is the method of having a dedicated resource room communication file go back and forth every time. There is the method of writing a short note and handing it over on the same day. There is the ...
note

If you are writing on "note", why not start small on Substack too? (It was a place where I could meet readers again and leave my "voice")

"Let's build your own base for sharing from your 50s." I write articles on note and Substack with that theme. The theme is "From your 50s...", but I am in my 60s. In reality, I am writing with this ...
GitHub

mmd-balls_as_credal_sets_a_pac-bayesian_framework_for_epistemic_uncertainty_in_t.md

The paper provides the first PAC-Bayes upper bound for test-time adaptation formulated as "Target Risk $\le$ Source Empirical Risk + KL Complexity + MMD Distribution Shift Term." It interprets ...
GitHub

a_language-guided_bayesian_optimization_for_efficient_lora_hyperparameter_search.md

description [ICML2026][模型压缩][LoRA调参] 本文把 LoRA 超参数配置写成带领域解释的文本，让冻结 LLM、可学习 token ...