目录

1. 介绍
2. 安装
3. 使用
   1. 示例
      1. 下载arxiv中最近的5篇cat:cs.AI类别论文。
      2. 下载arxiv中最近的5篇数学代数几何类别论文。
      3. 下载biorxiv中最近的2篇论文。
      4. 读取数据库。

介绍

下载并保存每日论文。

安装

python -m pip install auto-paper -U

使用

> python -m auto_paper --help

usage: auto_paper.py [-h] [--conf.src {arxiv,biorxiv,medrxiv}] [--conf.max-paper INT] [--conf.paper-db PATH] [--conf.pdf-dir PATH]
                     [--conf.query STR] [--conf.filters {all,new,today}] [--conf.sortby {Relevance,LastUpdatedDate,SubmittedDate}]
                     [--conf.filterby {updated,published}] [--conf.no-save-pdf] [--conf.no-show] [--url {None}|STR]

Run main function.

╭─ arguments ─────────────────────────────────────────────╮
│ -h, --help              show this help message and exit │
│ --url {None}|STR        (default: None)                 │
╰─────────────────────────────────────────────────────────╯
╭─ conf arguments ────────────────────────────────────────╮
│ Configuration.                                          │
│ ─────────────────────────────────────────────────────── │
│ --conf.src {arxiv,biorxiv,medrxiv}                      │
│                         (default: arxiv)                │
│ --conf.max-paper INT    (default: 100)                  │
│ --conf.paper-db PATH    (default: papers.db)            │
│ --conf.pdf-dir PATH     (default: pdfs)                 │
│ --conf.query STR        (default: cat:cs.AI)            │
│ --conf.filters {all,new,today}                          │
│                         (default: new)                  │
│ --conf.sortby {Relevance,LastUpdatedDate,SubmittedDate} │
│                         (default: LastUpdatedDate)      │
│ --conf.filterby {updated,published}                     │
│                         (default: updated)              │
│ --conf.no-save-pdf      (sets: save_pdf=False)          │
│ --conf.no-show          (sets: show=False)              │
╰─────────────────────────────────────────────────────────╯

示例

下载arxiv中最近的5篇cat:cs.AI类别论文。

python -m auto_paper --conf.max-paper 5

Fetching http://arxiv.org/pdf/2304.01196v1
Fetching http://arxiv.org/pdf/2202.01752v3
Fetching http://arxiv.org/pdf/2304.01179v1
Fetching http://arxiv.org/pdf/2304.01195v1
Fetching http://arxiv.org/pdf/2304.01201v1

> ls pdfs
2202.01752v3.pdf 2304.01179v1.pdf 2304.01195v1.pdf 2304.01196v1.pdf 2304.01201v1.pdf

下载arxiv中最近的5篇数学代数几何类别论文。

python -m auto_paper --conf.max-paper 5 --conf.query cat:math.AG

Fetching http://arxiv.org/pdf/2111.11216v3
Fetching http://arxiv.org/pdf/2304.01135v1
Fetching http://arxiv.org/pdf/2304.01149v1
Fetching http://arxiv.org/pdf/2101.12186v3
Fetching http://arxiv.org/pdf/2303.15776v2

> ls
2101.12186v3.pdf 2202.01752v3.pdf 2304.01135v1.pdf 2304.01179v1.pdf 2304.01196v1.pdf
2111.11216v3.pdf 2303.15776v2.pdf 2304.01149v1.pdf 2304.01195v1.pdf 2304.01201v1.pdf

下载biorxiv中最近的2篇论文。

python -m auto_paper --conf.max-paper 2 --conf.src biorxiv

Fetching https://www.biorxiv.org/content/10.1101/2021.01.11.426044.full.pdf
Fetching https://www.biorxiv.org/content/10.1101/2020.12.16.423137.full.pdf

> ls
10.1101.2020.12.16.423137.full.pdf 2111.11216v3.pdf                   2304.01135v1.pdf                   2304.01195v1.pdf
10.1101.2021.01.11.426044.full.pdf 2202.01752v3.pdf                   2304.01149v1.pdf                   2304.01196v1.pdf
2101.12186v3.pdf                   2303.15776v2.pdf                   2304.01179v1.pdf                   2304.01201v1.pdf

读取数据库。

import shelev

with shelve.open("papers.db") as db:
  keys = list(db.keys())

print(keys)
# ['http://arxiv.org/abs/2202.01752v3',
#  'http://arxiv.org/abs/2304.01195v1',
#  'http://arxiv.org/abs/2304.01201v1',
#  'https://www.biorxiv.org/content/10.1101/2020.12.16.423137',
#  'http://arxiv.org/abs/2304.01196v1',
#  'https://www.biorxiv.org/content/10.1101/2021.01.11.426044',
#  'http://arxiv.org/abs/2111.11216v3',
#  'http://arxiv.org/abs/2304.01179v1',
#  'http://arxiv.org/abs/2303.15776v2',
#  'http://arxiv.org/abs/2304.01135v1',
#  'http://arxiv.org/abs/2304.01149v1',
#  'http://arxiv.org/abs/2101.12186v3']

print(db["http://arxiv.org/abs/2202.01752v3"])
# {
#     'pid': 'http://arxiv.org/abs/2202.01752v3',
#     'title': 'Near-Optimal Learning of Extensive-Form Games with Imperfect Information',
#     'abstract': 'This paper resolves the open question of designing near-optimal algorithms\nfor learning imperfect-information extensive-form games
# from bandit feedback.\nWe present the first line of algorithms that require only\n$\\widetilde{\\mathcal{O}}((XA+YB)/\\varepsilon^2)$ episodes of
# play to find an\n$\\varepsilon$-approximate Nash equilibrium in two-player zero-sum games, where\n$X,Y$ are the number of information sets and $A,B$
# are the number of actions\nfor the two players. This improves upon the best known sample complexity
# of\n$\\widetilde{\\mathcal{O}}((X^2A+Y^2B)/\\varepsilon^2)$ by a factor of\n$\\widetilde{\\mathcal{O}}(\\max\\{X, Y\\})$, and matches the
# information-theoretic\nlower bound up to logarithmic factors. We achieve this sample complexity by two\nnew algorithms: Balanced Online Mirror
# Descent, and Balanced Counterfactual\nRegret Minimization. Both algorithms rely on novel approaches of integrating\n\\emph{balanced exploration
# policies} into their classical counterparts. We also\nextend our results to learning Coarse Correlated Equilibria in multi-player\ngeneral-sum
# games.',
#     'published': '2022-02-03',
#     'updated': '2023-04-03',
#     'categorie': ('cs.LG', 'cs.AI', 'cs.GT', 'stat.ML')
# }