代做DS 4023: Machine Learning Autumn 2024代做留学生SQL 程序

DS 4023: Machine Learning

Autumn 2024 Course Project

Description:

This  is  a  group  project   (3-5  group  members),  which  aims  at  applying  machine  learning algorithms (including but not limited to those covered in our lectures) to solve real-world tasks or conducting machine learning research.

Project Topics:

Each group will need to select a topic. There are two mainstreams of topics:

•    Application project: Pick  an application that interests you, and explore how best to apply learning algorithms to solve it.

•    Algorithmic project. Pick a problem or family of problems, and develop a new learning algorithm, or a novel variant of an existing algorithm, to solve it.

(some projects will also combine elements of applications, algorithms and theory)

You are suggested to pick something that you can get excited and passionate about, e.g., either an application area that you're interested in, or pick some subfield of machine learning that you want to explore more. For  inspiration, you might also look at some recent machine learning research papers. You can find research papers of top conferences in this  website https://aipapers-top.github.io/ , in which NeurIPS, ICML, ICLR are conferences most relevant to machine learning.  Alternatively,  if you're  already  working on a research or project that machine learning might apply to, then you may already have a great project idea.

To undertake the project, the following steps are essential:

1. Select one topic. (week 7-8)

2. Survey on existing research on relevant topics by searching related keywords on an academic search engine such as:http://scholar.google.com  (remember to narrow the topic to a feasible and suitable scope) (week 7-8)

3. Collect, read, and analyze relevant materials /data. (week 9)

•    An important aspect of designing your project is to identify one or  several datasets suitable  for  your  topic  of interest.  Get  the  benchmark  datasets  and  validate  your learning algorithms on the benchmark datasets is preferred. We don't want you to spend much time collecting raw data.

•    If you choose to use preprepared datasets (e.g. from Kaggle), we encourage you to do some data exploration and analysis to get familiar with the problem .

4. Design and implement learning algorithms/framework and validate the proposed algorithms (baselines) on benchmark/collected dataset. (week 10-11)

•    We expect a solid methodology, comprehensive validation and detail discussion of the experimental results.

•    Replicating the results in a paper can be a good way to learn. However, instead of just replicating a paper, also try using the technique on another application, or do some analysis of how each component of the model contributes to final performance.

5.    Produce the report and slides. (week 12)

A very good project report will be a publishable or nearly-publishable piece of written work. You may read some recent papers and follow the writing styles. For example, IEEE format is common  for  research  papers https://www.ieee.org/conferences/publishing/templates.html, https://www.overleaf.com/latex/templates/ieee-conference-template/grfzhhncsfqn .

Here     are     some     example      project     works     of     Stanford     University,     for     your reference:https://docs.google.com/spreadsheets/d/1GIsGZKozmTxqzmXLt7FBmVRCHVJvg d_1DD-f9f5PwkU/edit#gid=0,https://cs229.stanford.edu/proj2021spr/.

However, you should ensure your own topic because  these topics are relatively old (several years ago).

Submission Requirement:

Upon completion, each group must submit the following materials:

1. Project report, your report should contain but not limited to the following content:

a)    Motivation and background of the topic

b)    Related works and existing techniques of the topic (can use the content in the survey)

c)    Methodology

d)    Experimental study and result analysis

e)    Future work and conclusion

2. Link and description to the dataset and the implementation code.

3. Slides for presentation.

Submission deadline:

1.    Slides for presentation: Dec. 1, 2024

2.   Report and code: Dec. 20, 2024

Assessment:

In general, projects will be evaluated based on:

•    Significance. (Did the authors choose an interesting or a "real" problem to work on, or only a small “toy" problem? Is this work likely to be useful and/or have impact?)

•    The technical quality of the work. (i.e., Does the technical material make sense? Are the things tried reasonable? Are the proposed algorithms or applications clever and interesting? Do the student convey novel insight about the problem and/or algorithms?)

•    The novelty of the work. (Is this project applying a common technique to a well-studied problem (good performance), or is the problem or method relatively unexplored?)

The assessments are:

1.   Report: 30%

2.   Presentation: 50% (in the week 13-14, the detailed arrangement will be determined)

3.   Code: 20%