Day 1, R1 13:55–14:40 現今建立各種機器學習模型的套件很多，但是最常被詬病的是如何去解釋建立出來的機器學習模型。 尤其是若需要將模型應用到與人高度相關的行業中，如：金融業決定是否要發信用卡、醫療業協助診斷疾病等，若無法解釋模型，常會導致使用者或老闆不敢用、不相信模型之問題。 實際上現在已經有很多工具套件來協助解釋機器學習模型。 介紹三種常用的模型，以及如何運用套件來解釋這些模型。 包含線性迴歸可以用 p-value 和係數，Tree-based 模型可以用 feature importance 和 SHAP ，最後是類神經網路的模型一樣也可以用 SHAP 來對其做解釋。 主要會以結構化資料介紹 SHAP 套件的原理以及它的應用方式。 Slides not uploaded by the speaker. Speaker: Jack Pan 喜歡用Python開發的工程師。有三年以上的機器學習經驗，特別是將其運用到金融相關的問題上。 ... https://www.youtube.com/watch?v=bVsIIsObHss

PyCon Taiwan 2023｜Talk 演講｜Day 1, R2 13:05–13:35 ? 說明 Description ? Polars, the high-performance data manipulation library for Python and Rust, has been gaining traction as a promising alternative to Pandas for certain types of data analysis and manipulation tasks. With its optimized implementation of data structures and algorithms, Polars offers superior performance in joining, aggregating, and manipulating large datasets. If you're looking to revolutionize your data workflow, Polars may just be the ultimate replacement for Pandas. Whether you're working with big data, complex operations, or demanding computational tasks, Polars can help you get the job done faster and more efficiently. In this presentation, we'll take a closer look at the advantages of using Polars over Pandas, and explore some real-world examples of how Polars can improve your data workflow. We'll also discuss some of the key features and functionalities of Polars, including its support for multi-threading, columnar memory layout, and native integration with Rust. ? 投影片 Slides：https://slides.com/d/dFhlMlI/live ? 講者介紹 About Speaker - YVictor ? Specializing in Event-Driven Architecture for trading, my focus is designing and implementing systems that empower businesses in the financial industry. With expertise in event-driven messaging patterns, microservices, and workflows, I create architectures that enhance trading systems' speed, reliability, and scalability. I help organizations achieve higher agility by embracing event-driven principles and gaining a competitive advantage in fast-paced markets. Follow “PyCon Taiwan” ⭐️ Official Website: https://tw.pycon.org ⭐️ Facebook: https://www.facebook.com/pycontw ⭐️ Instagram: https://www.instagram.com/pycontw ⭐️ Twitter: https://twitter.com/PyConTW ⭐️ LinkedIn: https://www.linkedin.com/company/pycontw ⭐️ Blogger: https://conf.python.tw/ ... https://www.youtube.com/watch?v=QyW23zd2FXk

Day 2, R0 14:10–14:40 本講題將介紹深度機率語言(Deep Probabilistic Programming)的核心概念與特性，並分享如何使用Pyro來體驗深度機率語言有趣與強大之處。 Pyro為基於Pytorch之上所開發的深度機率語言(Deep Probabilistic Programming)，能建立複雜機率模型並使用隨機變分估計(Stochastic Variational Inference)方法於GPU上進行大規模訓練，本講題會分享實作文本分析LDA模型、變分自編碼器VAE心得。 Slides: https://docs.google.com/presentation/d/1qOIqK5MmE-b43yTYgQL7b7Opu-AmkqGGCgmvgdHhqfg/edit?usp=sharing Speaker: 柯維然 穿上制服是個菜鳥警官，忙於勤務與警政工作；脫下制服是個熱愛用數據與程式解決問題的資料戰警；目前服務於新竹市警察局，並於交通大學資訊工程研究所攻讀博士班，主要研究方向為深度生成模型、機率模型等 技術部落格：https://odie2630463.github.io ... https://www.youtube.com/watch?v=HKIjenYy4qw

PyCon Taiwan 2016｜一般演講 Talks ? 摘要 Abstract ? 一個平常沒在做數學相關研究的工程師，因緣際會下使用到了 SymPy。藉機來介紹 SymPy 的基本使用，從建立自己的方程式到使用 SymPy 所支援的方程式化簡，最後利用 Codegen 轉換到其他程式語言或 Library，並檢視藏在這些功能背後的 AST (Abstract Syntax Tree)。 ? 關於講者 About Speaker - Chiu-Hsiang Hsu ? 社會新鮮人，去年剛畢業、當完兵，目前正在新創公司工作，我的朋友大多叫我 "dv"，同事都叫我 "CH"，目前活動範圍以桃園、台北居多，熱愛 Open Source :) #python #pycontw #pycontw2016 Follow “PyCon Taiwan” ⭐️ Official Website: https://tw.pycon.org ⭐️ Facebook: https://www.facebook.com/pycontw ⭐️ Instagram: https://www.instagram.com/pycontw ⭐️ Twitter: https://twitter.com/PyConTW ⭐️ LinkedIn: https://www.linkedin.com/company/pycontw ⭐️ Blogger: https://pycontw.blogspot.com ... https://www.youtube.com/watch?v=2CQyOutb7aI