There are 66 posts in total till now.
Long Range Arena: A Benchmark For Efficient TransformersConference: ICLRYear: 2021link: https://openreview.net/pdf?id=qVyeW-grC2k This is the new benc
ICLR 2017 Smoothing對NLP來說,很容易發生test dataset的字並沒有出現在training dataset,所計算 c(word) 會等於零。這不代表那些字就不重要,所以要用Smoothing的方法調整機率。 If data sparsity isn’t a prob
paper : Attention Is All You Need AbstractThe dominant sequence transduction models are based on complex recurrent or convolutional neural networks th
TSVM (Transductive Inference for Text Classification using Support Vector Machines)tags: SVM, semi-supervised learning, SSLpaper linkcited:3517參考書籍:Le
前情提要: 為甚麼需要git?如果不用版本控制工具 就會有以下的悲劇…. GIT要建專案時若要用git來追蹤,則強烈建議先用github 建專案後在clone到本地端(你的電腦)否則之後若要把在本地端建立的專案 傳到github 時會變得比較麻煩 初始設置(從github clone 下來的不用做)
Data Science and Machine Intelligence Lab (DSMI Lab) locates in SA314 at National Chiao Tung University. Our researches focus on data science, machine