Effective Approaches to Attention-based Neural Machine Translation

tags: study paper DSMI lab

paper: Effective Approaches to Attention-based Neural Machine Translation

Introduction

  • Neural Machine Translation (NMT) requires minimal domain knowledge and is conceptually simple
  • NMT generalizes very well to very long word sequences => don’t need to store phrase tables
  • The concept of “attention”: learn alignments between different modalities
    • image caption generation task: visual features of a picture v.s. text description
    • speech recognition task: speech frames v.s. text
  • Proposed method: novel types of attention- based models
    • global approach
    • local approach

Read More

Leetcode - Two Sum

在面試時,如果順利解完白板題了,面試官可能會接著問:

“這樣的時間複雜度是多少?”

“還能更快嗎?”

所以,我們試著從簡單的問題來學習如何分析時間複雜度xD

Time complexity


試圖用個不嚴謹的方法來理解他,想像有一個計數變數count = 0在你的程式裡

每執行一次你的程式count就會+1,請問執行完後count會等於多少?

Read More

Introduction of time series models

tags: references

本篇文章介紹常用的時間序列model和相關概念,希望以後大家有遇到時序分析的問題的時候,稍微知道他們在幹嘛XD

Notations:

$r_t$: a time series, $a_t$: a white noise series, $\rho_l$: lag-$l$ autocorrelation,
$\gamma_l:Cov(r_t,r_{t-l})$,$\sigma^2$: variance of $a_t$, $B$: back-shift operator

名詞解釋:

  • Stationary:
    • Strict: distribution is time-invariant (基本上不可能達到)
    • Weak: first 2 moments are time-invariant (平均、標準差、共變數不隨著時間變)

Read More

Efficient Estimation of Word Representations in Vector Space

論文網址: Efficient Estimation of Word Representations in Vector Space

4 model architectures:

NNLM; RNNLM; CBOW; Skip-Gram

為了比較模型好壞,先定義接下來訓練深度模型的複雜度皆為:$$O = E\times T\times Q$$

  • E: 迭代次數
  • T: 訓練集的詞個數
  • Q: 模型參數

old: NNLM, RNNLM

  1. What’s NNLM (Feedforward Neural Net Model)?


    (上圖是原始paper(A Neural Probabilistic Language Model)的圖,annotation可能會跟下面word2vec的對不起來,下面使用原本word2vec的annotation)

有4層: input, projection, hidden, output

Read More

Overview of Meta Learning

Meta Learning

Introduction

Meta Learning = Learn to learn 讓機器學習如何去做學習這件事。

實際例子:
當我們接觸到一個裝糖果的玻璃罐時,我們察覺玻璃罐與保特瓶相似的本質,因而有辦法套用既往的知識快速的移轉到新的任務上,而Meta learning便是在學這個過程,在遍覽多種任務後,學習一組對任務敏感的參數,當新任務進來時能快速的將先驗知識移轉到新任務中。

Read More


Powered by Hexo and Hexo-theme-hiker

Copyright © 2020 - 2021 DSMI Lab's website All Rights Reserved.

UV : | PV :