Reviewing (how NOT to review a paper)  Paper Reviews  Goals: paper selection, ensure high quality, constructive feedback and recommendations, widen own horizon  Lots of similarities to code reviews Drafts  #3 Mindset: Quality over Quantity  Aim for top-tier conferences/journals (act as filter)  Make the paper useful for others (ideas, evidence, code)  #4 Easily Readable: Quality ∝ Time  Make it

Includes submitted paper and final presentation (pass := adhere to given constraints + acceptable quality) Course Organization 10 706.015 Introduction to Scientific Writing – 01 Introduction and Overview (matplotlib), Gnuplot  Figures: MS Visio, Inkscape  pdf, eps, svg (vector graphics)  #3 Mindset: Quality over Quantity  Aim for top-tier conferences/journals (act as filter)  Make the paper useful for University of Technology, WS 2020/21 Writing the Paper (and more Experiments)  Easily Readable: Quality ∝ Time  Make it easy to skim the paper  paragraph labels, self-explanatory figures (close to

collection of peer-reviewed, free, open- source C++ libraries. Boost code is generally very high-quality, is widely portable, and fills many important gaps in the C++ standard library, such as type traits that a type au- thor usually doesn’t have, and shouldn’t need. The stakes here are high because low-quality hash functions can be security vulnerabilities, due to the emergence of hash flooding attacks. Even compound types, because the implementation cannot recur- sively call std::hash on data members. High-quality hash algorithms maintain large amounts of internal state, and reducing that state to the size_t bytes

preprocessing, ResNet18, climate models  Label Creation/ Validation  Team learning  Labeling w/ checks  Label validation  Quantitative validation w/ 10 expert votes on correctness Reproducibility