What is the position of Kaggle in the MachineLearning field?

What kind of people will win the kaggle award? Can I learn something? Is it a highlight of job hunting or postgraduate application when I receive a prize at kaggle? What kind of people will win the kaggle award? Can I learn something? Is it a highlight of job hunting or postgraduate application when I receive a prize at kaggle?

Reply content:

From a year or two after Netflix Prize, this type of competition has become prevalent. At the beginning, some of the experts who participated in the promotion were excellent. For example, Xiang Liang, who was well-known, was a second member of Netflix Prize.

Over the past few years, as the number of competitions has increased, winner solution has become increasingly popular. No matter what the game is, LR + GDBT + FM + NN, then ensemble will always get a good result.

What rankings are obtained on kaggle cannot be described. If the competition shows the ability to analyze and solve problems, especially the ability to propose targeted results, it can reflect the real level. For example, the Shanghai Jiao Tong University APEX lab team developed the SVD Feature after attending KDD Cup 11; some tree-related technologies involved in KDD Cup 12 won the first place are the foundation of Tianqi's ICML 13 paper and XGBoost. On the contrary, if you open some solutions for the recent kaggle competition, most of them follow some specific procedures, and there is nothing new.

In addition, now everyone is in a group. The only few of them can really make a decision. In addition, there are also a few who like to play soy sauce with their own scores, so they should be more careful with these people.

Participating in the competition is different from doing research. Many people do not have to understand the details of model and algorithm. They can achieve good results by running a few open-source packages. * Good * research requires a deep understanding of models and applications. Therefore, it is a bright spot for job seeking, but not necessarily for graduate students. The winners are all people with real skills. You can earn a good ranking on several questions above, which proves your practice and understanding in the field of data science. I am recruiting almost the following standards:

1. I have participated in the Kaggle competition. I will read my resume.
   

2. I will call for an interview once at 10%.

3. I will give an on site interview if I have two or more 10% requests.

4. We have to talk about it in the first 10 days.

The landlord cheers. In the mainstream scientific research field, the impact may not be big, but it is still very useful in the industrial field. If you have some special highlights, it will be very convincing. As mentioned in @ lau phunter.

If you want to get a good score in the Kaggle competition, you will inevitably need to do a lot of experiments: About parameter selection, model selection, and Feature Engineering. In order to efficiently complete these experiments, we must have good experimental ideas and solid code skills to complete the Pipeline design and architecture. This is a test of the comprehensive capabilities of people and is required by the industry.

Even so, the Kaggle competition is much simpler than the real-world machine learning. Where have we spent all our time?
* Determine the problem to be solved.
* Clear optimization indicators
* Collect appropriate data
* Data cleansing
* Conduct various experiments
* Other groups are required to perform A/B Test together.
* Integrate the machine learning Pipeline into the Pipeline of other products
* Selling our models on various occasions is really useful ......

So there is not much time to really run the experiment. However, at least one good result in Kaggle indicates that you can perform experiments systematically, which is a very big highlight and a very good indicator. It should be useless to apply for a graduate student. I do not know if I have applied for a job using this item. As for whether you can learn something, see what kind of competition, and whether you are willing to work hard.

Some of the competition data is too simple. download and run an Xgboost, and it will be 10%. However, if you try new things on your own, you will also be rewarded.

Some of the data is troublesome, and it takes time to process the data. The feature engineering space is also rich. Sometimes you have to write some rules or write the loss function on your own. You can learn a lot from these competitions.

It is not easy to win. In a simple competition, because there are too many participants and the method is too homogeneous, the time is spent on tuning parameter and ensemble. It takes good luck to win the prize. In a complex competition, the methods vary widely. If we still have the tuning parameter Set, it will not work. The top team's time is spent on the key to the problem. Whoever gets the key is more, and whoever gets the score is higher, it takes a lot of time and thinking. What kind of people are there. There are many "professional players" from large companies in competitions with high prizes. Most of them are entry-level players with low prizes. Looking for a job in a company that lacks professional machine learning talents will help. It is better to apply for a graduate student to post a top article. I feel that I have no position, and it is a powerful tool to become a professional and become a data scientist. This kind of practical competition is quite good. We have done a big data competition by imitating kaggle. Welcome to play.
Yuan prize, time hacker: Find the programmers who have created the time to use the cloud calendar product Big Data Mining & online programming CompetitionOne suggestion, do a crappy PhD, or get a job. just don't get a master's. unless it's funded or in US. it is not difficult to enter the list. It is easier to rank first, and it is difficult to rank first. It is a routine, and it is good to be proficient + simple thinking. It's no different from an excavator's wine bottle opening competition...

What is the position of the Kaggle competition in the Machine Learning field? For more information, see the PHP Chinese website (www.php1.cn )!