Both aspects of the students, the research direction is AI machine learning algorithm engineer, or do after to master relevant knowledge based machine learning, must grasp the search skills, especially research or working direction in the field of the latest paper, more advanced skills, is this paper, the algorithm can emersion This is a solution when the authors of the paper don’t have the open source code, but until we can master the skill, we want to be able to search for code reproduced by others.
Therefore, today I would like to recommend two related websites, both of which are the result of the same team. The two websites, one can be used to show the paper algorithm with code implementation, and the other gives the latest and best algorithm paper results in various fields.
1. Papers with Code
First give the url of the website:
paperswithcode.com
This is Reddit a user rstoj do a website, the latest machine learning on the ArXiv papers and lot code (TensorFlow/PyTorch/MXNet/etc.). Compared to the previous recommended reading of ArXiv’s website, this user made the most of the need to satisfy more researchers — to find the code to implement the thesis algorithm!
The project indexes about 50,000 papers (published on ArXiv in the last five years) and 10,000 on Github.
You can search by keyword by title, or by keyword by field of study, such as image classification, text classification, etc. You can also rank by popularity, latest papers, and most Stars on Github. This site will keep you up to date with what’s going on in the machine learning community.
First, take a look at what the site might look like:
The figure above shows the paper in order of popularity, giving the title of each paper, the author, the conference or summit to which the paper was submitted, some introduction, such as whether it is the most advanced algorithm in the field, tags (i.e. Keywords, the direction of the paper), and the code framework adopted (e.g. Pytorch, TensorFlow, or other frameworks). Links to the paper and code, as well as the current number of Stars on Github and the number of stars being added per hour.
In order of the highest number of Github stars, it looks like this:
You can see that the first two stars are TensorFlow, and the third is Caffe frame.
In addition, if we enter keywords in the search box for the research area, such as Image Classification, the search results will look like this:
It shows a total of 250 current papers containing papers in the field, then shows the best algorithms and papers on several data sets, and open source projects (if open source), then subdomains -Subtasks, and finally the papers with code in that domain, ranked by number of Github stars.
2. Browse state-of-the-art
Again, give the url first:
paperswithcode.com/sota
This site is mainly to solve another problem – to find a field of the best (State of the art, Sota) algorithm papers and implementation code. It is just come into contact with a new field, must want to do, the best algorithm to find the latest paper first, and then according to the code realization, this paper first run down the code, and then to get to know the details, or is according to its reference papers, to study in this field in recent years papers (usually 3 years to 5 years), Gradually get familiar with the research direction and difficulties in this field.
Again, the team at Papers With Code has created a website that can query the latest algorithms in the domain, with a total of 16 categories, 950+ individual subclass tasks, 500+ evaluation results (including Sota results), 700+ databases, and 8000+ Papers. As shown in the figure below:
The 16 categories include:
- Computer vision
- Natural language processing
- medical
- The research methods
- miscellaneous
- voice
- The game
- Figure (Graphs)
- The time series
- audio
- The robot
- music
- reasoning
- Computer coding
- The knowledge base
- Antagonism (Adversarial)
Click on the computer vision category, and you can see that it is divided into 450+ sub-tasks, as shown in the figure below:
Here you can continue to click on each subtask, such as image classification, and you will get the following picture:
The picture above is actually a legend of the keyword examples in the search field when we introduced the Paper with Code website.
Finally, give the urls of the two websites:
- paperswithcode.com
- paperswithcode.com/sota
With the help of the above two websites, I believe it can help you better to learn and be familiar with the work of the research field direction!
Welcome to follow my wechat official account – Machine Learning and computer vision, or scan the qr code below, we communicate, learn and progress!
Past excellent recommendation
Learning notes
- Introduction to Machine Learning series 1 — An Overview of machine learning
- Introduction to Machine Learning (2)– How to Build a Complete Machine Learning Project
- Acquisition of machine learning data sets and construction of test sets
- [GAN Learning Series] First acquainted with GAN
- [GAN Learning Series 2] The origin of GAN
- [GAN Learning Series 3] Image Restoration using Deep Learning and TensorFlow (PART 1)
Python-100 Exercise series
- Python – 100 | exercises 01 & list comprehensions
- Python-100 Exercise 02
Math study notes
- Programmer’s Math note 1– Base conversion
- Programmer’s math note 2– remainder
- Programmer’s Math Note 3– Iterative methods
Github Project & Resource Tutorial recommended
- [Github project recommended] a better site to read and find papers
- TensorFlow (TensorFlow
- Must read AI and deep learning blog
- [Tutorial] A simple and easy to understand TensorFlow tutorial
- [Resources] Recommend some Python books and tutorials, both beginners and advanced!