Machine learning in labs and companies
Focus on applications
The most common pitfall that I see with companies and labs is to focus on methods. This will get you only so far. Here is why, and what you can do different to avoid wasting time and resources.
Machine learning difficult?
If you think that machine learning is difficult then you are right, but probably for the wrong reasons.
Many perceive machine learning as difficult because of the math (and maybe coding) that underpins it. What they do not realize is that, in order to make fruitful use of machine learning, you do not need to understand this math.
Expertise can be bought and implementations can be found often for fee.
The hard part is to figure out you what machine learning can do for your research or for your business. This questions needs deliberate thought. This makes machine learning hard, and not the math.
Develop an understanding
In order to apply machine learning successfully, you need to develop an understanding of what machine learning can do for you. That allows you to assess how different choices can impact your research or business. It enables you to identify opportunities and to see risks. Building this understanding is key to success.
It requires effort, yet it is important not to skip this phase. I dare go as far to state that that skipping this step is one of the main reasons why many machine learning projects fail.
Once you have figured what machine learning can do for you, hiring someone to implement your ideas is a relative small step. You will feel much more comfortable and you will have much higher chances of success.
The good news
Figuring out what machine learning can do for you can be a lot of fun. It is a creative phase were you will review what competitors are doing and where will articulate a direction that you want to go.
How do I figure out what machine learning can do for me?
It is less hard than you think. In one of my next posts I will write about a format that I use with clients and that that everyone can use.
Meanwhile, let me know your thoughts on this topic!
Working on a simulation or measurement pipeline that is too slow to iterate on? Book Free Consultation.