Participating in a coding interview has to be one of the most intimidating aspects of conducting a job search as a new software developer. You have to pull yourself away from the projects you are working on, and dedicate some of that time to stretching your problem solving skills, and understanding of data structures and algorithms, just to have a shot at an offer. Problem solving can be a lot of fun, and is a source of enjoyment for many in the field, but when you have a potential offer riding on your ability to develop an optimal solution, the pressure can be immense.
My first coding challenge was a couple months ago, just as I finished bootcamp and was trying to wrap my head around heaps and Dijkstra’s algorithm. The position was for new-graduates who would be placed in a rotational program in the software team for this social media company. I felt it was a great fit for my entry level skills, and was fortunate to have my resume accepted. Then came the coding challenge. I received an email containing a link to complete an automated test within the next 3 days. I spent the first 2 blowing through an algorithm and data structure course online, hoping any of the information would stick and help me solve the problems. I hadn’t put in the time I needed in order to feel prepared to tackle the challenges.
Once I started the test I was relieved to see the first 2 questions weren’t complicated. They involved using binary search and multiple pointers, fairly easy skills to pick up and understand. Then came the third and final problem. It was a version of the classic algorithm that I had yet to encounter, the ‘Weighted Job Scheduling’ problem. The problem goes, given an array of jobs, each containing a start time, and end time, and a profit, determine the maximum profit you can obtain from completing non-overlapping jobs. The problem can be arranged in many different ways. I have seen it described as ‘2 arrays, the first array, n, containing start times of presentations, and the other array, m, containing the length of presentations, where n[i] relates to m[i]’ and you are tasked to find the maximum presentations that can take place. The input could be one or two arrays, or an object, containing more arrays or objects with keys. For my problem, the input could look like the following:
jobs = [[1,2,15],[1,3,20],[2,4,40],[3,5,75],[4,5,50],[1,4,35]]
jobs[i][0] = start-time
jobs[i][1] = end-time
jobs[i][2] = profit
So, having not practiced much dynamic programming, I was naive of the complexity of the problem, and how to solve it. I tried to brute force my way to a solution, given the first test case, but each test case proved to break my solution as the input grew and became unsorted. Long story short, I failed the challenge and did not progress in the process.
This is a problem looking for an optimal solution, and it wasn’t until I understood how to use tabulation, that I was able to crack the solution. Looking back at that input, if we try to find the optimal solution while we go through the array, it can look like this:
jobs = [[1,2,15]]
optimal solution is jobs[0] as its the only option
jobs = [[1,2,15],[1,3,20]]
optimal solution is jobs[1] as jobs[1][2] > jobs[0][2] and they conflict
jobs = [[1,2,15],[1,3,20],[2,4,40]]
now [jobs[0],jobs[2]] is the optimal solution as they don't conflict, and generate a larger combined profit than jobs[1].
jobs = [[1,2,15],[1,3,20],[2,4,40],[3,5,75]]
but now [jobs[1],jobs[3]] are optimal
As you can see, every time we consider an additional input, we have to put that in context of what we have already seen, and reevaluate what is optimal, which seems to be an complex task when dealing with large data sets. How can we keep track of every potential solution as we take in more data? The answer is tabulation.
#algorithms #software-development #software-engineering #coding #flatiron-school