YoungForest's blog

After the National Day holiday, Second Senior Brother and I had a chance to go to Guangdong. We visited Guangzhou, Macau, and Hong Kong in order. On October 6, we set off from Beijing, took a 10-hour high-speed train to Guangzhou South Railway Station, and then checked into Guangdong Hotel.

Guangzhou

On the first day, we rested and spent the day in Guangzhou. We got up early to climb Yuexiu Park. Since it was the last day of the long holiday, young people were all tired from playing and lying at home to rest. On the streets and in the park, we could only see large numbers of middle-aged and elderly people exercising. In the park, I finally saw with my own eyes the “kapok” trees that grow in the south. It reminded me of the lines from To the Oak: “I wish to be a kapok beside you…” and of the plot in the TV drama To the Oak. Oaks and kapok trees grow in the north and south respectively, and are destined not to grow together. Only by knowing this can one understand the sorrow conveyed in To the Oak.

How should I enjoy the last holiday of 2018, National Day? I chose to join a group tour, Beijing Chuxin Outdoor, to go horseback riding on the Bashang Grassland in Fengning, Hebei. At 19:30 on the evening of September 30, we gathered at the south entrance of Huitong West Street and got on the bus. We headed north. By the time we reached a service area at the northernmost part of Beijing around 9:30, it already felt cold. When we arrived at a farmhouse lodging in Fengning County near midnight, the weather was already very cold. According to the host, it had just snowed the day before, so these two days were especially cold. Later, after National Day, when scrolling through Moments, I saw that people who went to deserts, Gobi, or volcanoes were all cold without exception. So in autumn, going to these places really requires careful thought and full preparation. Before departure, the tour leader repeatedly emphasized bringing thick clothes, preferably a down jacket. I have been afraid of cold since childhood, so I directly brought my thickest down jacket, a Canada Goose with 500g of down. It is said that this jacket can even handle Antarctica. I have never been to Antarctica, so I do not know whether that is true. But it had no problem resisting the temperature of an autumn grassland night.

Looking back on four years of university: classes, competitions, friendships, change, and pivotal choices.

Yesterday I participated in an intern interview with JingChi Technology. This internship interview was arranged uniformly by the lab, not something I found myself.

The format was online interview. There were two rounds in total, each expected to last one hour, but the second round actually only asked questions for half an hour. The first round used Skype. The call quality was poor, and there was no video output. The second round used WeChat voice plus collabedit, and the effect was much better.

The technical interviews were both very simple. After all, they were hiring coding interns for a data annotation platform, so naturally the requirements were extremely low. I felt like I was about to be sold cheaply.

Description: https://leetcode.com/problems/kth-largest-element-in-an-array/description/
Solution: https://leetcode.com/problems/kth-largest-element-in-an-array/discuss/
Difficulty: Medium

This was a problem my senior schoolmate was asked in his interview with JingChi. His interview was in the morning, and mine was in the afternoon. So after talking with him about the interview content, I solved all the problems he had been asked, including this one and Coin Change.

My Dynamic Programming

From August 21 to August 28, I spent seven days traveling in Shandong with a feeling of self-exile. When I was little, in 2008, I once followed my father and a tour group to Shandong. We visited the classic tourist cities, including Qingdao, Yantai, Weihai, and Dalian across the sea. I was still young and innocent then, and only felt that not having to go to school and traveling to unfamiliar places was very interesting. Now ten years have passed. The scenic spots were the same, but my mood and feelings were different.

This trip was decided on temporarily. Earlier, classmates had invited me to Chongqing for a graduation trip, but for various reasons I did not manage to go. After mid-August, after some emotional turmoil and after being brainwashed by Sanmao’s wandering literature, a classmate from Shandong, whom we can call “Female-Voice Guy” because his voice is especially feminine, invited me to travel around Shandong together. I agreed. I hoped this trip would make me feel a little better, filling my mind with a hurried journey so I would not think about troubles I should not think about.

Description: https://leetcode.com/problems/convert-bst-to-greater-tree/description/
Solution: https://leetcode.com/problems/convert-bst-to-greater-tree/solution/
Difficulty: Easy

Although this problem is Easy, writing a correct solution in one pass is still difficult.
When you see a binary tree, you should think of solving it recursively. One trick in this problem is how to pass along the value that needs to be accumulated during recursion.

When recursively calling the right subtree, we need to return the sum of the entire right subtree and add that sum to the root node. In my specific solution, the right subtree’s sum may exist in two places: 1. the leftmost leaf node of the right subtree; 2. the root node of the right subtree, when the right subtree has no left child.

Description: https://leetcode.com/problems/single-number/description/
Solution: https://leetcode.com/problems/single-number/solution/
Difficulty: Easy

The difficulty of the problem lies in this requirement: Your algorithm should have a linear runtime complexity. Could you implement it without using extra memory?

I thought hard for a long time but still could not satisfy both O(n) time complexity and O(1) space complexity. I went to read the solution, and Approach 4 meets the requirement. It uses the properties of the XOR bit operation, which really is a bit of a trick. You can also see the comment section full of exclamations like “awesome.” Once you know it, it is not hard; the next time I encounter it, it will be Easy.

A Tsinghua graduate was the interviewer. At the beginning, he asked me about my future work intentions. I answered “algorithm engineer,” but in fact I was interviewing for a development role, so the senior person lectured me a bit. He asked whether I had applied to the wrong group and whether I still had not found the direction I was interested in. Then he asked what I was doing in my current internship at Kuaishou. I talked about some of it, and then got criticized again. He said this feature engineering work would all be replaced by deep learning later.

When the formal interview started, he only asked two questions: implementing a queue that maintains the minimum value, and implementing a stack that maintains the minimum value.

There was supposed to be a second round afterward, but later they could not find an interviewer, so it fizzled out.

To train my data mining ability more quickly, I plan to spend some time on Kaggle competitions over the next month. The current plan is to follow the competitions, Datasets, and expert notebooks under Tutorial as an introduction. After that, I will choose a simple regular competition to participate in.
As for whether I will eventually follow Brother Huan onto the path of data mining competitions, that depends on how I feel after getting started and whether I am willing to continue spending a lot of energy on it.

Data mining ability is mainly divided into two parts:

• machine learning
• feature engineering

When I previously interned in the recommendation team at Kuaishou, the work also involved these topics. Recommendation systems can also be considered a branch of data mining. In terms of output, they predict whether a user will click a certain item.

Titanic: Machine Learning from Disaster, as most people’s first step into Kaggle, really is a concentrated showcase of data mining skills. By contrast, the competitions that have recently been popular on Kaggle are mostly related to CV or NLP. They require more domain knowledge and dilute the importance of general data mining skills.
There are also many other tutorials on Kaggle for users to study. In the future I will choose some of them for focused learning and try to get started with Kaggle before the semester begins.