More often than not, aspiring researchers tend to waste a lot of time with their research, especially when it comes to machine learning. They ponder around to find an unachievable idea or drag the process of carrying out their machine learning research for too long. So, below are given some of the best habits for budding machine learning researchers to strategise the process.
First and foremost, before getting into these, one must have a good understanding of the machine learning concepts and fundamental principles. These best habits mentioned below are not technical but are a guide towards how one can save on time when thinking about starting a machine learning research.
Sign up for your weekly dose of what's up in emerging technology.
Search for an Interesting and a Fun Problem to Solve
Whatever technical skills one has developed over time, they best come out when one puts them to the test on real-world problems or instead on the type of problems that are fun and exciting to them. If one’s aspiration is to make a breakthrough, they should be working towards an idea that is interesting to them and has enough applications in the real world.
And searching for an idea is usually about searching for something that stands out to you. One has to think about what aspect of machine learning piques their interest, what new results have been achieved in the field, and how this result changes their view on the technology.
These ideas or research topics don’t have to be the original ones. One can define a unique problem by asking different questions to existing ones. To carry out some form of research, one should be able to produce deep insights, build algorithms, and develop results of a problem or idea they are researching on.
Deciding on the Problems
Now that one has a clear idea about what kind of problem they would want to do research on, they have to develop a particular taste in it. One has to continually keep asking questions and must always be going through similar ideas as time goes on. Over time, one develops a distinct sense of what kind of ideas will prevail and what kind of ideas don’t have the potential to manifest into detailed research.
So, what ‘always going through the ideas’ means? It means that one has to go through as many papers and books related to the problem as possible. Not only read through them but also assess them to find a way to get an acute understanding of the subject and also try to indulge in discussions around the topic with experts.
Seek advice from experts and ask questions to yourself related to the topic like, how is this theory useful? When can these results be transferable to the real-world? What causes a particular idea to take a broad uptake? Why are some of these forgotten?
One of the reasons big ideas make it into the real-world is that the work is tightly clustered to a small number of researchers. Big ideas come out because of their expertise and the environment they are working on. They might have more knowledge and a bigger perspective, but it does not necessarily mean that one working in a less dense environment will not have a significant impact, they just have to work a little harder.
Developing an idea for a particular sector and implementing it can be daunting at times. One has to have expertise in the area and should have the capabilities of developing solutions for eventually finding a way to make the project even better from their initial observations.
But the problem of coming up with a unique project is that there are chances that someone else might also have developed the same idea. Suppose whatever publication or paper one is reading, the same one is available for everyone around the world and someone might follow the same line of thought as you.
Now, developing an original idea is difficult, drawing out ideas or developing a different opinion from an existing study is one of the more comfortable options when it comes to deciding what to work on. One has a clear understanding of what different angles can be taken from the usual approach and also it gives a distinctive approach to one’s study with a more precise and imaginable goal.
An end-goal that drives an idea generally channels one’s team towards tackling different aspects of the project and makes it easier to go through the ups and downs because they will have an idea of what the result will be like.
But, there is a problem with this approach, the problem of narrowing one’s research too much. What happens with this is that, whatever results achieved with the project cannot be applied to other aspects of machine learning, thus stalling the advancements in the field.
Do Not Get Carried Away With The Goal
Many times the problem with failing to do proper research or the best ones is that people often do not work on solving the important problems.
Before starting on a problem, people need to ask themselves about the various aspects of the problem. They need to ask themselves questions about its potential, the success rate etc.
One of the most significant factors that come into play when setting high goals is the fear of being unsuccessful. Now, for example, if you have an idea about training a robot to do a specific task, say, running and performing acrobatic stunts like backflips. This goal is something that expert researchers have been working on for years now. So, if you are someone who possesses enough knowledge to make a robot hold something steadily, making it do stunts will be a huge goal to achieve and often could be demotivating to follow through.
One needs to tackle the smaller important problems and then work towards the bigger ones. Targeting the smaller problems will reduce the intimidation factor and will also help with self-developing. Also, while working on a problem, emphasis must be given to the amount of improvement it will yield and also the complexity of the problem.
Now while one is doing research, they have to set apart some time for their personal development. During the research, one can find many other exciting ideas, encounter new challenges and acquire new skills. One has to set some time apart for themselves to improve their knowledge about machine learning in general. What happens with carrying out with one’s daily work is that, over time their understanding of the research limits them from getting new information and this area becomes a comfort zone. It is a piece of common knowledge how dangerous a comfort zone is.
Making notes simply means documenting every piece of information about an idea and jotting down new insights into a notebook. In a more relatable context, you might have experienced forgetting ideas that pop up in your head when you are performing some other tasks or during the research and fail to remember them later on. A notebook or a space on your device could be somewhere you can write them down.
Not only is the notebook space for your ideas, but also a single stop where you can find all the results related to your ongoing research.
Some Tips Towards Understanding ML Algorithms and Research Better
- Write the models from scratch: To truly understand, write a model from scratch
- Build testable and assertive code: Write codes which are easy to understand and friendly to others. This will help one to write working models faster
- Try as many tools and platforms
- Implement Ideas from other research papers
- Build a training system
- Run the model
- Record everything you have done
- Practice with other unusual problems