How to Choose Machine Learning or Deep Learning for Your Business
Business + ManagementDeep LearningMachine LearningAccelerate AIDeep LearningMachine Learningposted by Elizabeth Wallace, ODSC May 2, 2019 Elizabeth Wallace, ODSC
AI is the future, or so you’re hearing. Every day, news of another organization leveraging AI to produce business outcomes that outstrip competition hit your inbox, but your company either hasn’t started at all or is mired in the discussion.
AI, machine learning, and deep learning are sometimes used interchangeably, but they aren’t the same. If your business is going to leverage advances in technology, you need to know the difference and when to choose machine learning over deep learning and vice versa.
[Related article: 5 Roadblocks to Getting an ML System in Production]
Machine Learning Versus Deep Learning: What’s the Difference?
Short story: deep learning is a subset of machine learning and both fall under the umbrella of AI. A lot of it has to do with the type of data you have. Structured data can train machine learning algorithms but deep learning does best with unstructured data. So which one do you have? You have to identify the predominant source of data before deploying, or you could end up with an expensive mess.
Remember the different types of tests you took in school? Multiple choice, true/false, essay, all those. Structured data is a lot like true/false or multiple choice. There’s a single answer for the question. If we are grading those tests and you ask me “What’s the answer to number one?” I can easily reply “False.” No other explanation or context is needed.
If we want to train someone else to grade the test, all we need is a table. Simple columns and rows and each block filled in with correlating, relational information. Number one is “False.” Number two is “true” and so on. Someone could learn to grade the test quickly.
Structured data is information easily accessible and processed. It’s clean and offers no variance or context. There’s no need. Each column and row is filled in, so your algorithms require far fewer training sets to get the right answers. The upside? It’s all clear. The downside? Just like true false doesn’t give us much insight into what our students know, running structured data offers surface-level insights.
We are still grading exams, but we move on to long form, essay-based exams. When you ask me what the answer is to number one, the answer isn’t so clear. Essays have a lot of factors that go into a “right” answer and students can answer any number of ways and still be correct.
Training someone to grade these tests requires significantly more investment. Creating a guide isn’t possible with our table method because there are so many factors to consider. Not every student has the same information. Multiple types of information are possible. Partial credit is available. At best, we create a rubric and have someone grade many exams with us to help “norm” grading.
Unstructured data is like this. It’s messy, unprocessed, and filled with variance. The upside? Just like it’s much harder to “guess” the correct answer and therefore tells us so much more about our students’ comprehension, working with unstructured data gives businesses a richer, deeper insight into the area they’re questioning. The downside? It’s complicated to train and difficult to assess.
[Related article: 5 Things Business Leaders Should Know about Data Science]
Choosing Machine Learning Or Deep Learning
Machine learning is a lot like our exam trainee in the true/false test. There’s no need to spend countless hours training the examiner to handle true false. The answers are clear and given enough tests, the examiner could even memorize the answers to grade faster.
If your business is dealing with a lot of structured data, i.e., email addresses, gender, identity, etc., there may be no need to go through deep learning algorithms. Machine learning can verify identities, respond to simple queries, and offer fundamental insights into the gender or age of customers buying a particular product, for example.
Machine Learning is best for:
- classifying information
- regression models or predicting outcomes based on past performance
- clustering or placing like information together
Deep Learning should happen if your business is dealing with much more massive amounts of data (mind-bogglingly large) and not all that data is processed. If a company needs to figure out how people feel about a particular product that, despite positive user testing, is tanking, or predict future trends with a current product, deep learning can offer much better insight into what’s going on.
If you’re asking questions too complex for basic machine learning or all your data is unstructured (and can’t be cleaned), deep learning could be the way to go. You must have the data, however, on a scale too complex for humans to sort.
Deep learning is best for:
- large scale analysis of unstructured data (doctor’s notes, for example)
- uncovering hidden elements
- enterprise analysis
Choosing Machine Learning or Deep Learning
Smaller businesses can still leverage machine learning to offer value to their customers provide insight that can help drive business decisions. Having a data scientist on board (or a data science team) can be an excellent way to help decide which type of program to launch for your particular needs. Although machine learning isn’t quite as complex as deep learning, machine learning offers relevant insights for more straightforward data. There’s no need to overcomplicate things.
Your business will benefit from leveraging either of these models. Your market and your competition are moving towards using machines to offer insights into the data available, so put a plan in place and take the first step.
Want to learn more about choosing ML or DL for your business in-person? Check out the Ai x Business Summit at ODSC East 2020 this April and learn more about implementing AI into your organization today.