原创 2019.11.5 note
2019.11.5 note Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer They propose a fr
原创 2019.11.15 note (2)
2019.11.15 note (2) Intriguing properties of neural networks DeepFool: a simple and accurate method to fool deep
原创 2019.11.6 note
2019.11.6 note DeepGCNs: Making GCNs Go as Deep as CNNs Graph Convolutional Networks (GCNs) offer an alternative th
原创 2019.11.15 note (1)
2019.11.15 note (1) EXPLAINING AND HARNESSING ADVERSARIAL EXAMPLES FGSM (Fast Gradient Sign Method) Adversarial
原创 IMO 2017 T1解答
Puzzle欄目開通 生活中經常做到有意思的題目,自己也會有些巧妙的解法。當時不記下來很快會忘記,非常可惜。所以我在博客裏開了Puzzle這個欄目,記錄有意思的謎題。那麼這個欄目的第一題,就獻給今天早上剛剛出爐的IMO 2017吧
原创 2019.12.17 note
2019.12.17 note Filter Response Normalization Layer: Eliminating Batch Dependence in the Training of Deep Neural Ne
原创 計算學習理論基礎
計算學習理論基礎 基礎知識 與計算理論類似,機器學習也有PAC可辨識,PAC可學習,高效PAC可學習的概念。 有限假設空間 在有限假設空間情形下,可以對PAC學習的樣本複雜度在可分,不可分,不可知情形下做出估計。 VC維
原创 2020.1.13 note
2020.1.13 note AdderNet: Do We Really Need Multiplications in Deep Learning? Compared with cheap addition operation
原创 VC維的幾個習題
VC維的幾個習題 線性分類假設空間的VC維 VCH(Rd)=d+1 ,H爲線性假設空間。 這是西瓜書上的一個習題,在學習計算理論的時候我也在博客上記過這個結論,二維的線性空間VC維爲3,(4的時候異或類的標號就打不散)也十分顯然
原创 2019.12.13 note
2019.12.13 note Fast Transformer Decoding: One Write-Head is All You Need They have proposed an alternative to
原创 2019.10.18 note
2019.10.18 note Quaternion Knowledge Graph Embeddings In this work, authors move beyond the traditional complex-val
原创 2019.10.8 note
2019.10.8 note Commonsense Knowledge Base Completion This work proposes a task of complete commonsense knowledge ba
原创 2019.10.12 note
2019.10.12 note Exploiting Explicit Paths for Multi-hop Reading Comprehension The task and proposed method for 2 ho
原创 2019.9.5 note
2019.9.5 note A Structural Probe for Finding Syntax in Word Representations The probe identifies a linear transfor
原创 2019.9.4 note
2019.9.4 note A Simple Theoretical Model of Importance for Summarization Define Redundancy, Relevance and Informa