masa8 / paper

Note for research papers interested in

Write your paper

How to begin writing?

1. Install TexShop on your mac
2. Open TexShop
3. Write your paper
4. Command + T

How to start writing using default layout

1. cp -r tex_base
2. Write a peper with default layout!

How to add References?

1. Command + T
2. Command + B
3. Command + T

How to update References?

Arxiv Example

1. Visit Arxiv : https://arxiv.org/abs/2211.03679
2. Check Cite as: arXiv:2211.03679
3. Visit arxiv2bibtex: https://arxiv2bibtex.org
4. Enter 2211.03679
5. Copy cite and Save it
6. Open BibDesk
7. Select Library
8. Drop the saved file

ScienceDirect Example

1. Visit ScienceDirect : https://www.sciencedirect.com/science/article/abs/pii/S016786552200085X
2. Click "Cite" and "Export citation to BibTeX"
3. Open BibDesk
4. Select Library
5. Drop the downloaded file

Ref.

This is awesome: http://gottscott.github.io/2016/08/30/how-to-do-citations-in-latex.html

paper's Issues

Snuba: Automating Weak Supervision to Label Training Data

INPUT

Paroma Varma Stanford University
[email protected]
http://www.vldb.org/pvldb/vol12/p223-varma.pdf
https://github.com/HazyResearch/reef

OUTCOME

• Solve this bottleneck
  • a key bottleneck is gathering enough high-quality training labels tailored to each task

OUTPUT

• Snuba

  • Snuba automatically generates heuristics in under five minutes and performs up to 9.74 F1 points better than the best known user-defined heuristics developed over many days.
  • Snuba outperforms other automated approaches like semi- supervised learning by up to 14.35 F1 points.

• Overview

  • the synthesizer (Section 3.1) that generates a candidate set of heuristics, a pruner (Section 3.2) that selects a heuristic to add to an existing committed set of heuristics, and a verifier (Section 3.3) that assigns probabilistic labels to the data and passes the subset of the labeled data that received low confidence labels to the synthesizer for the next iteration.

• Synthesizer
  The Snuba synthesizer takes as input the labeled set, or a subset of the labeled set after the first iteration, and outputs a candidate set of heuristics

- Decision Stumps
- Logistic Regressor
- KNN

• Pruner
  
• Verifier
  

Multi-views Embedding for Cattle Re-identification

INPUT

Luca Bergamini
[email protected]
https://arxiv.org/pdf/1902.04886.pdf

OUTCOME

Person Re-ID works a lot actively, but for animal, it is not. I want to change this.

OUTPUT

• Method

  • Multi-view network
  • Trained with Histogram Loss
  • Triplet loss didn't work
    

• Dataset

  • Train Set; consisting of 12952 pictures from 387 different subjects;
  • Database Set; consisting of 4289 pictures from 52 differ-ent subjects, recorded during two different days;
  • Test Set; consisting of 561 pictures from 52 different subjects.
    These cows are the same included in the Database Set;

• Result
  
  We present extend baseline comparisons both with non-deep and deep methods