Difference between revisions of "Step 1"

From Research management course
Jump to: navigation, search
 
(28 intermediate revisions by the same user not shown)
Line 1: Line 1:
−
The most important things come first. We discuss the main message, delivered by a scientific paper. We explore the first three elements of a scientific paper: the abstract, the highlights, and the keywords. The main message shall reveal itself through all elements of the paper. But we leave the rest of it for the next time. Namely, the title, introduction, problem statement, goal of the computational experiment, and conclusion are left behind. We select a part and exercise in the reconstruction of these three elements.  
+
The most important things come first. We discuss the main message, delivered by a scientific paper. We explore the first three elements of a scientific paper: the abstract, the highlights, and the keywords. The main message shall reveal itself through all elements of the paper. But we leave the rest of it for the next time. Namely, the title, introduction, problem statement, goal of the computational experiment, and conclusion are left behind. We select a paper and exercise in the reconstruction of these three elements.  
  
 
== The seminar ==
 
== The seminar ==
−
# [https://m1p.org/index.php?title=Step_1 The warm-up 3-minute test]  
+
# [https://forms.gle/FrUzQbRSLPTVRMXM9 The warm-up 3-minute test]  
 
# Model, Algorithm, Method: Machine learning in a nut-shell  
 
# Model, Algorithm, Method: Machine learning in a nut-shell  
 
<!-- #* more terms: statistical hypothesis, algebraic structure, model selection, bayesian inference -->
 
<!-- #* more terms: statistical hypothesis, algebraic structure, model selection, bayesian inference -->
−
# Step 0 homework results discussion
+
# Step 1 homework, how to read: the scheme <!--(how to search is a separate topic)-->
−
# Step 1 homework, how to read (how to search is a separate topic)
+
<!-- # Structure of the main message -->
−
# Structure of the main message
 
 
# Structure of the abstract
 
# Structure of the abstract
−
# The second and the last slide of your talk
+
<!-- # The second and the last slide of your talk -->
 
# Extracting keywords
 
# Extracting keywords
 
# Highlights: compressing the paper
 
# Highlights: compressing the paper
Line 16: Line 15:
 
#* LaTeX: compile your file and commit without temporary files
 
#* LaTeX: compile your file and commit without temporary files
 
# The papers to select from
 
# The papers to select from
 +
# Step 0 homework results discussion
 
# Optional GPT-role discussion
 
# Optional GPT-role discussion
  
 
==Resources==
 
==Resources==
−
Step 1 Youtube video (expected with online version)
+
Step 1 YouTube [https://youtube.com/live/EZH3RdSXRtc video]
 +
'''Warning!''' A wrong microphone was used. This video will be rewritten in a couple of days.
 
<!--  
 
<!--  
 
* [https://youtu.be/5RVkgUOYiro Step 1 Youtube video]
 
* [https://youtu.be/5RVkgUOYiro Step 1 Youtube video]
Line 27: Line 28:
  
 
==Homework==
 
==Homework==
−
# Set up your GitHub repository using the template, [https://docs.github.com/en/repositories/creating-and-managing-repositories/creating-a-repository-from-a-template see how]
+
# Set up your GitHub repository using [https://github.com/vadim-vic/the-Art-homework/ this template], see [https://docs.github.com/en/repositories/creating-and-managing-repositories/creating-a-repository-from-a-template how]
 
# Select a paper to read from the list below
 
# Select a paper to read from the list below
 
# Reconstruct its
 
# Reconstruct its
Line 33: Line 34:
 
## Keywords
 
## Keywords
 
## Highlights
 
## Highlights
−
## Short motivation for why you selected this paper (no templates here it is an extra topic to discuss)
+
## Short motivation for why you selected this paper (no templates here, since it is an extra topic to discuss)
 
# Compile and upload TEX and PDF to GitHub (no temporary files, please)
 
# Compile and upload TEX and PDF to GitHub (no temporary files, please)
−
# Linear models. Read pages XX from the book XX, for the next warm-up test
+
# Fill out the [https://forms.gle/KqhRk9R6w61snAB9A Step 1 questionnaire]
 +
# Refresh in your memory the Linear models for the next warm-up test, either 
 +
## look for the terms [https://en.wikipedia.org/wiki/Dot_product dot product], [https://en.wikipedia.org/wiki/Scalar_projection scalar projection], [https://en.wikipedia.org/wiki/Linear_least_squares least squares], [https://en.wikipedia.org/wiki/Transformation_matrix linear map]
 +
## or do fun-reading, the pages 33-39 from [https://klassfeldtheorie.wordpress.com/wp-content/uploads/2018/10/mathematische-methoden-310117.pdf the book] Section L3.
  
 
'''Note''' that we always respect your credit hours. So please keep track of it.
 
'''Note''' that we always respect your credit hours. So please keep track of it.
  
 
'''Your profit''' here is your ability to find the main message of a paper.
 
'''Your profit''' here is your ability to find the main message of a paper.
 +
 +
=== How to read ===
 +
There are many pieces of advice on how to read scientific papers, see [https://forums.fast.ai/t/how-to-read-research-papers-andrew-ng/66892 an example].
 +
<!-- including [https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7392212/ exhaustive ones].-->
 +
We suggest briefly looking through the paper's
 +
# highlight, or pitch in the abstract,
 +
# central formulas,
 +
# clarifying figure,
 +
# plots and tables,
 +
# find the main idea.
 +
And questions, what are:
 +
# the topic?
 +
# the subject of research?
 +
# the main idea or message?
 +
# the impact, is it useful for you?
 +
 +
===Abstract===
 +
The abstract of a paper is the first piece the reader looks at. Usually, it is written at the beginning of research and after the paper is done, before submission. Due to its importance, several versions of the abstract from different points of view are welcome.
 +
 +
The abstract is limited to 600 characters. It may contain
 +
# wide-range field of the investigated problem,
 +
# narrow problem to focus on,
 +
# features and conditions of the problem,
 +
# the idea of the suggested solution,
 +
# the novelty and alternative solutions to compare with,
 +
# application to illustrate with.
 +
 +
Examples of abstracts to discuss, [https://m1p.org/images/d/db/M1p_2024_lect2_c.pdf a draft].
 +
<!-- [http://www.machinelearning.ru/wiki/images/1/19/TheSecondSlide.pdf Think of a motivation of your research in slides]. -->
 +
 +
===Keywords===
 +
The keywords of your paper shall match the subject of your research, and show the area and the focus. Ensure these keywords are used in your paper frequently and play an important role. They shall be recognized terms in your field of knowledge. See detailed [https://www.redwoodink.com/resources/how-to-choose-the-best-keywords-for-your-research-manuscript explanations] and Elsevier [https://scientific-publishing.webshop.elsevier.com/manuscript-preparation/how-choose-keywords-manuscript/ recommendations].
 +
 +
===Highlights===
 +
To write highlights, see [https://www.elsevier.com/researcher/author/tools-and-resources/highlights elsevier] official version, a useful piece of
 +
[https://editingindia.wordpress.com/2015/07/14/writing-highlights-for-elsevier-dos-and-donts/ advice], and [https://medium.com/@miguel_93656/writing-meaningful-highlights-in-scientific-papers-4371ff33ab8a Medium] clarifications.
  
 
==Papers to choose from==  
 
==Papers to choose from==  
 
Please read a paper from this list and formulate its main message. Imagine you are a journal editor or a reliever, who receives scientific papers randomly and pick up some paper.  
 
Please read a paper from this list and formulate its main message. Imagine you are a journal editor or a reliever, who receives scientific papers randomly and pick up some paper.  
−
#  
+
 
−
#  
+
''If these papers are too difficult to you to understand'', there is no big deal. Most likely, you were going to read a paper of your own interest. Read it. The main requirements, it must be a scientific paper. See the next section.
−
#  
+
 
−
#  
+
You can briefly go through the bold items of [https://cseweb.ucsd.edu/~wgg/CSE210/howtoread.html How to Read an Engineering Research Paper by W.G. Griswold]
−
#  
+
 
−
#  
+
'''IMPORTANT'''. Since the homework is to reconstruct the abstract of one of these papers, please, try to skip the published abstract. Cover it and start reading according to the discussed reading scheme.
−
#  
+
 
−
#  
+
# Distinguishing time-delayed causal interactions using convergent cross mapping [https://doi.org/10.1038/srep14750 DOI]
−
#  
+
# Comprehensive study of feature selection methods to solve multicollinearity problem according to evaluation criteria [https://doi.org/10.1016/j.eswa.2017.01.048 DOI], [https://m1p.org/papers/Katrutsa2016QPFeatureSelection.pdf PDF]
−
#  
+
# Spatio-temporal filling of missing points in geophysical data sets [https://doi.org/10.5194/npg-13-151-2006 DOI]
 +
# Analytic and stochastic methods of structure parameter estimation [https://doi.org/10.15388/Informatica.2016.102 DOI]
 +
# Longitudinal predictive modeling of tau progression along the structural connectome [https://doi.org/10.1016/j.neuroimage.2021.118126 DOI]
 +
# Generative or Discriminative? Getting the Best of Both Worlds [https://www.microsoft.com/en-us/research/wp-content/uploads/2016/05/Bishop-Valencia-07.pdf PDF]
 +
# Neural Ordinary Differential Equations [https://proceedings.neurips.cc/paper_files/paper/2018/file/69386f6bb1dfed68692a24c8686939b9-Paper.pdf NIPS], [https://arxiv.org/pdf/1806.07366 Appendix]
 +
# Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations [https://doi.org/10.1016/j.jcp.2018.10.045 DOI], [https://github.com/maziarraissi/PINNs GitHub]
 +
# How much does it help to know what she knows you know? An agent-based simulation study [http://dx.doi.org/10.1016/j.artint.2013.05.004 DOI]
 +
# GRAND: Graph Neural Diffusion [https://proceedings.mlr.press/v139/chamberlain21a/chamberlain21a.pdf PMLR]
  
 
===Can I select a paper of my own choice?===
 
===Can I select a paper of my own choice?===
 
Yes. Here are some formal requirements.  
 
Yes. Here are some formal requirements.  
−
# A clear message in the area of Theoretical Foundations of Machine Learning.
+
 
 +
# A clear message in the area of Machine Learning.
 
# No Kaggle-style papers with messages like "It works, but nobody knows how".  
 
# No Kaggle-style papers with messages like "It works, but nobody knows how".  
 
# Top peer-reviewed journals, no ArXiv, better avoid conferences.
 
# Top peer-reviewed journals, no ArXiv, better avoid conferences.
−
# No papers from other fields: linguistics, medicine, finance, physics, etc. There must be only one primary subject: Machine Learning.
+
# No papers from other fields: linguistics, medicine, finance, physics, etc.  
 
# No overviews of paper collections, it is another genre.  
 
# No overviews of paper collections, it is another genre.  
 
# No [https://en.wikipedia.org/wiki/Predatory_publishing predatory publishing houses]
 
# No [https://en.wikipedia.org/wiki/Predatory_publishing predatory publishing houses]
 +
 +
In this case please write an explanatory text about why you choose this paper.
  
 
=== Recommended journals ===
 
=== Recommended journals ===

Latest revision as of 00:47, 22 September 2024

The most important things come first. We discuss the main message, delivered by a scientific paper. We explore the first three elements of a scientific paper: the abstract, the highlights, and the keywords. The main message shall reveal itself through all elements of the paper. But we leave the rest of it for the next time. Namely, the title, introduction, problem statement, goal of the computational experiment, and conclusion are left behind. We select a paper and exercise in the reconstruction of these three elements.

The seminar

  1. The warm-up 3-minute test
  2. Model, Algorithm, Method: Machine learning in a nut-shell
  3. Step 1 homework, how to read: the scheme
  4. Structure of the abstract
  5. Extracting keywords
  6. Highlights: compressing the paper
  7. Instastructure for your homework
    • GitHub: organize the repository
    • LaTeX: compile your file and commit without temporary files
  8. The papers to select from
  9. Step 0 homework results discussion
  10. Optional GPT-role discussion

Resources

Step 1 YouTube video Warning! A wrong microphone was used. This video will be rewritten in a couple of days.

Homework

  1. Set up your GitHub repository using this template, see how
  2. Select a paper to read from the list below
  3. Reconstruct its
    1. Abstract
    2. Keywords
    3. Highlights
    4. Short motivation for why you selected this paper (no templates here, since it is an extra topic to discuss)
  4. Compile and upload TEX and PDF to GitHub (no temporary files, please)
  5. Fill out the Step 1 questionnaire
  6. Refresh in your memory the Linear models for the next warm-up test, either
    1. look for the terms dot product, scalar projection, least squares, linear map
    2. or do fun-reading, the pages 33-39 from the book Section L3.

Note that we always respect your credit hours. So please keep track of it.

Your profit here is your ability to find the main message of a paper.

How to read

There are many pieces of advice on how to read scientific papers, see an example. We suggest briefly looking through the paper's

  1. highlight, or pitch in the abstract,
  2. central formulas,
  3. clarifying figure,
  4. plots and tables,
  5. find the main idea.

And questions, what are:

  1. the topic?
  2. the subject of research?
  3. the main idea or message?
  4. the impact, is it useful for you?

Abstract

The abstract of a paper is the first piece the reader looks at. Usually, it is written at the beginning of research and after the paper is done, before submission. Due to its importance, several versions of the abstract from different points of view are welcome.

The abstract is limited to 600 characters. It may contain

  1. wide-range field of the investigated problem,
  2. narrow problem to focus on,
  3. features and conditions of the problem,
  4. the idea of the suggested solution,
  5. the novelty and alternative solutions to compare with,
  6. application to illustrate with.

Examples of abstracts to discuss, a draft.

Keywords

The keywords of your paper shall match the subject of your research, and show the area and the focus. Ensure these keywords are used in your paper frequently and play an important role. They shall be recognized terms in your field of knowledge. See detailed explanations and Elsevier recommendations.

Highlights

To write highlights, see elsevier official version, a useful piece of advice, and Medium clarifications.

Papers to choose from

Please read a paper from this list and formulate its main message. Imagine you are a journal editor or a reliever, who receives scientific papers randomly and pick up some paper.

If these papers are too difficult to you to understand, there is no big deal. Most likely, you were going to read a paper of your own interest. Read it. The main requirements, it must be a scientific paper. See the next section.

You can briefly go through the bold items of How to Read an Engineering Research Paper by W.G. Griswold

IMPORTANT. Since the homework is to reconstruct the abstract of one of these papers, please, try to skip the published abstract. Cover it and start reading according to the discussed reading scheme.

  1. Distinguishing time-delayed causal interactions using convergent cross mapping DOI
  2. Comprehensive study of feature selection methods to solve multicollinearity problem according to evaluation criteria DOI, PDF
  3. Spatio-temporal filling of missing points in geophysical data sets DOI
  4. Analytic and stochastic methods of structure parameter estimation DOI
  5. Longitudinal predictive modeling of tau progression along the structural connectome DOI
  6. Generative or Discriminative? Getting the Best of Both Worlds PDF
  7. Neural Ordinary Differential Equations NIPS, Appendix
  8. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations DOI, GitHub
  9. How much does it help to know what she knows you know? An agent-based simulation study DOI
  10. GRAND: Graph Neural Diffusion PMLR

Can I select a paper of my own choice?

Yes. Here are some formal requirements.

  1. A clear message in the area of Machine Learning.
  2. No Kaggle-style papers with messages like "It works, but nobody knows how".
  3. Top peer-reviewed journals, no ArXiv, better avoid conferences.
  4. No papers from other fields: linguistics, medicine, finance, physics, etc.
  5. No overviews of paper collections, it is another genre.
  6. No predatory publishing houses

In this case please write an explanatory text about why you choose this paper.

Recommended journals

  1. Machine Learning
  2. Expert Systems with Applications
  3. Journal of Machine Learning Research
  4. Artificial Intelligence
  5. Neurocomputing
  6. IEEE Transactions on Pattern Analysis and Machine Intelligence
  7. Neural Networks
  8. Pattern Recognition
  9. Data Mining and Knowledge Discovery
  10. Nature Machine Inlelligence, the problem is the first word here is Nature so it focuses on natural sciences

See also

  1. Elsevier's open archive
  2. Springer's open archive


Transcript of the video

Appears after the seminar.

Retrieved from ‘https://m1p.org/index.php?title=Step_1&oldid=1766’