4 Reasons to Start Participating in Data Science HackathonsDenis VorotyntsevBlockedUnblockFollowFollowingMay 3My hackathon participation path started about a year and a half ago.
I managed to take part in more than 20 various sizes and topics events, which took place in Moscow, Helsinki, Berlin, Munich, Amsterdam, Zurich and Paris during the current time.
In all these events I’ve been engaged in data analysis in various forms.
I like visiting new cities, establish new contacts, come up with fresh ideas, quickly implement old ones fast as well as feeling adrenaline during a speech and results announcement.
The current post is the first of three ones, devoted to hackathon topics.
I’m going to tell you about hackathon concept, as well as the reason to participate in hackathons.
The second post will show the organizer’ mistakes and their further results.
The third post will provide answers to the most popular questions about hackathons.
What is a hackathon?Hackathon lasts several days and its goal is to resolve certain issues.
Usually, hackathon resolves several issues, each of which is a separate track.
The sponsoring company provides the task description, success metrics (metrics can be either subjective, like «novelty and creativity» or objective — classification accuracy on the deferred dataset) and resources to achieve success (API, hardware, datasets).
Participants should formulate the issues, propose a solution and show their product prototype for a certain time.
The best solutions get both awards as well as further cooperation opportunity.
Hackathon stagesOnce the tasks are announced, hackathon participants are united into teams: each «loner» gets a microphone and talks about the chosen task, his background, idea as well as the specialists he needs to cooperate with in order to implement it.
Sometimes a team may consist of just one person who is able to perform all the work independently at a high enough level.
This is relevant for data analysis hackathons, but it is often forbidden or not recommended for product events — the event organizers are aimed at further project development continuation in the company; the previously formed team has got several advantages over those participants, wishing to create a product alone.
The optimal team members amount is 4 persons, namely a front-end, a backend, a Data Scientist as well as a businessman.
The optimal team members amount is 4 persons, namely a Front-end, a Backend, a Data Scientist as well as a Businessman.
By the way, data science/product hackathons separation is quite simple — data science hackathon implies there is a dataset available with a clear metric and leaderboard or there is an opportunity to win with the code in the Jupiter notebook while product one implies all the rest, namely it’s required to make an app, website or something attractive.
Usually, the project starts at 9 pm on Friday while its deadline is 10 am on Sunday.
Certain time should be taken to sleep (sleep is mandatory, coding 30 hours straight is a sure way to failure, I checked it myself), which means that the participants have got too little time to develop a highly-quality product.
The company’s representatives and mentors are available to help the participants.
Sleep is mandatory, coding 30 hours straight is a sure way to failure, I checked it myself.
Project operation starts with communication with the company’s representatives since they better realize the task specifics, metrics and most likely they will judge your completed work.
The current communication goal is to realize which areas are the most relevant as well as the main aspects to focus on.
One hackathon set the following task: regression on a dataset with tabular data, pictures as well as a clear metric — RMSE.
Once I talked to the company’s Data Scientist, I realized they needed a classification, but not a regression — the fact is someone decided that the best issues solution way wat that one.
Besides they require classification to realize which parameters are the most important while making a decision and then processing them manually, but not to obtain the monetary metric's growth.
I.
e.
the initial task (regression with RMSE) changes to classification; the assessment priority changes from the accuracy obtained to the result explanation opportunity.
This denies the opportunity to use stacking and black box algorithms.
The simple dialogue saved me a lot of my time as well as increased my chances to win.
The simple dialogue saved me a lot of my time as well as increased my chances to win.
Be sure to proceed to work once you realize the task clearly.
You are required to set the checkpoints — the time which it’s required to complete the tasks by; besides it’s a good idea to continue communicating with mentors — both company’s representatives and technical specialists since this is useful to timely make corrections during project development.
Taking a fresh look at the issue may suggest an interesting solution.
The organizers used to give both lectures and master classes since lots of newcomers participate in hackathons.
The lectures amount is 3, as usual, namely the lecture on own idea presentation way in the product form, a lecture on technical topics (for example, open APIs usage in machine learning to avoid own speech2text development for two days, but use the existing ones instead), a lecture on pitching (the way to present own product, How to use own hands when public speaking to avoid the audience bored).
There are various ways to encourage participants, namely to practice yoga, table football and tennis or a gaming console.
There are various ways to encourage participants, namely to practice yoga, table football and tennis or a gaming console.
You are required to provide your operation results to the jury on Sunday morning.
Good hackathons imply all to start with technical expertise whether something you claim really works?.The current test goal is to separate teams who have got just beautiful presentations instead of real products from the guys who really managed to develop something.
Unfortunately, not all hackathons imply technical expertise which sometimes results in those teams winning who have got over 12 slides presentation instead of a real product.
Such precedents are not frequent, but since they are best remembered, which results in the fact many people are sure a good presentation is a 99% victory in hackathons.
The fact is a presentation, is rather important, but its contribution is no more than 30%.
The fact is a presentation, is rather important, but its contribution is no more than 30%.
The jury makes a decision on awarding the winners once the participants speaking is completed.
So, the official hackathon part is completed.
Motivation to participate in hackathons1.
ExperienceHackathon is a really unique event in terms of experience gained.
In fact, there are just a few places to implement own idea for just 2 days and get instant feedback.
Critical thinking, teamwork skills, time management, the ability to work in a stressful situation, to present own operation results in an understandable form, presentations skills and lots of others are being improved during the hackathon.
Therefore, hackathons are a great place for people with theoretical knowledge to get really valuable experience.
2.
PrizesTypically, the hackathon prize fund is approximately 1.
5k — 10k euros for the first place (in Russia — 100–300 thousand rubles).
The expected participation value can be calculated while using a simple formula:EV = Prize * WinRate + FutureValue — Costswhere Prizeis the prize size (we’ll assume there is just one prize);WinRate— victory probability (the current value for a novice team will be limited to 10% on top, for the one more experienced — over 50%; I’ve met people who left each hackathon with an award, but this is rather an exception to the rules since their win rate will be below 100%);FutureValueis a one which demonstrates the future participation profit: it can either profit in the acquired experience form or established connections, information received, etc.
The current value is almost impossible to determine exactly, but it must be remembered;Costs— transportation costs, accommodation costs, etc.
The decision of whether to participate is made on hackathon EV with activity EV comparing basis, which you would like to do instead of hackathon: if you were going to lie all day long and watch TV, then you should be involved in the hackathon; but your plans were to spend time with your parents or a girlfriend, then may become your team for a hackathon (just kidding, decide yourself), if you were going to freelance — just compare the dollar-hour.
According to my own calculations, I can definitely say that participation in Russian hackathons is commensurate with the monetary profit from the usual working day for the average junior — middle level data scientist, but there are some important features (team size, hackathon type, prize fund, etc.
).
So, today hackathons may provide a pleasant increase to your personal budget, but not are a really golden opportunity.
3.
Work at the company and networkingHackathon is one of the ways to hire new employees for companies.
It’s a great way to show you are an adequate person and are aware of the way to work on hackathon in contradistinction to an interview while twirling a binary tree on a board (which doesn’t mean it correspond to the tasks you are going to complete while working, but the traditions must be honored).
Such «combat» conditions test may replace the whole test day.
I’ve got my first job due to participation in hackathon.
I showed a great opportunity to get more money out of the data as well as the way I was going to do that.
So, I won and then continued to implement that project while working in the sponsoring company.
That was the fourth hackathon in my life.
4.
Unique datasetsThis is a rather relevant point for the data science hackathons, but unfortunately, not everyone realizes its importance.
Usually, sponsoring companies provide real datasets during the event.
The current data is private while being protected by NDA, which doesn’t prevent to show the proof of concept on a real dataset instead of a toy titanic.
Such results will greatly help to apply for a job in a given company, a competitor company or in justifying similar projects in the future.
It’s obvious, all other things being equal, it’s great to have completed projects which have got positive feedback.
In general, such implemented projects play a similar role with medals and statuses on Kaggle, but their value for the industry is more obvious.
TipsParticipation in a hackathon is a rather diverse background and it is difficult to formulate the rules list.
However, here I’d like to provide the observations list, able to help a newcomer:Feel free to participate in hackathons even if you don’t have a background or team.
Think about what the benefits to bring.
Maybe you have an interesting idea or are you well versed in a certain field?.You have a great opportunity to use your domain knowledge while formulating an issue as well as to find non-trivial solutions.
Or maybe you can google the best?.Your skill will save a lot of time if you can find ready-made implementations in a GitHub.
Or are you very good at Lightgbm parameters tuning?.In such a case be sure to prove it in Kaggle competitions instead of participating in hackathons.
Tactics are more important than maneuvers.
Hackathon sets the task to resolve the issues.
Sometimes it’s required to reveal the issue in order to resolve it.
Check whether your the issue identified is really relevant for the company.
Be sure to check your solution for compliance with the issue and ask yourself about your solution optimality.
The issue of urgency and the proposed solution adequacy are the most important while evaluating your decision.
Your neural network architecture doesn’t matter.
Visit as many hackathons as possible, but feel free to leave badly organized events.
Add your hackathon operation results to your resume and feel free to write about it publicly.
Very true for hackathons Minimum Viable Product (MVP).. More details