紀錄怎麼用 OpenCV 產生影片—VideoWriter、fourcc 編碼器、sorted、glob。
紀錄從影片中提取影像—VideoCapture、read()、frame 遍歷、isdir、print。
紀錄用 `dirent.h` 計算資料夾檔案數量 — 用 opendir、readdir 遍歷,`d_type == DT_REG` 判斷一般檔案。
30 天 Google ML 鐵人賽的最終回顧 — 把 coursera 上《How Google does Machine Learning》與《Launching into Machine Learning》兩門課的學習路徑整理成索引,以及參賽心得、未來計畫與感謝。
Lesson 13 — 用混淆矩陣(confusion matrix)分析 ML 模型好壞,以及三個可評估的指標:Accuracy、Precision、Recall。
Lesson 12 — 完美的損失函數存在嗎?介紹 ML 模型的性能指標(performance metrics),以及不適當的最小 loss 帶來的問題。
Lesson 11 — 怎麼判斷 ML 模型該停止訓練了 — 用驗證資料集、測試資料集、交叉驗證(cross-validation)評估模型表現。
Lesson 10 — 泛化(generalization):怎麼檢查 ML 模型對未知資料的預測能力,以及 underfitting / overfitting 的問題。
Lab 5 — 在視覺化網頁中觀察 neural network 怎麼分類資料,並比較 batch size 如何影響 gradient descent 的收斂。
Lab 4 — 用 TensorFlow Playground 在視覺化網頁中體驗訓練 ML model,觀察 learning rate 改變對訓練過程的影響。