Guest blog by Eric Brown, Technology consultant, investor, entrepreneur
In a meeting with a CIO and her team recently, I mentioned the term “deep learning” in the context of some big data and machine learning initiatives the CIO had asked me to investigate. This particular CIO is fairly savvy when it comes to big data, machine learning, and data analysis, but she stopped me mid-sentence to ask me to explain what I meant when I used the term “deep learning.” We spent the next one-and-a-half hours walking through the basics of artificial intelligence, machine learning, and how this organization could incorporate these great approaches into their business.
With the growth of big data and data science, I come across a lot of questions and discussions about machine learning, but I rarely come across discussions about deep learning and the value it can bring to an organization.
From machine learning to deep learning
First, it’s important to understand the basis for deep learning. Machine learning is the foundation for deep learning and can be described as giving a computer a set of instructions that allow it to “learn” more about the subject it’s being asked to analyze. As an example of machine learning, you really don’t need to look any further than your email client and the spam filters that are used to filter out junk mail. Generally, these spam filters use a form of machine learning to learn what “good” and “bad” emails look like, and then they filter your email inbox based on what the machine learning algorithm has learned about the emails you receive.
Deep learning isn’t quite as simple and easy to understand as a spam filter. Deep learning is a branch of machine learning that attempts to build models based on different datasets that allow computer systems to understand and learn from the underlying abstracted data, in order to make more informed decisions about new datasets that come into the deep learning system. The advantage of deep learning systems is that there’s no need to “train” the system based on a preconceived notion of what a dataset might represent.
Applications of deep learning
I’ve seen (and built) deep learning systems in many fields, including telecommunications, finance, security, social media, insurance, and the automotive industry. One example that touches just about all of us is the use of deep learning systems to monitor credit or debit card activity to catch fraudulent use before it gets out of control. Another example that many readers will have interacted with is Netflix’s use of deep learning within their recommendation engine to improve recommendations to users on what they should watch based on what they’ve watched and liked in the past.
One last example that might help add context to the use of deep learning and the type of results it can deliver is AdTheorent, which has used a deep learning system to help advertisers target and engage with mobile users. This particular system is built to deliver real-time bidding assistance to advertisers bidding on ads for mobile devices. Using a number of datasets, this deep learning system helped advertisers reach an engagement level that was 200 to 300 percent higher than the industry average for the 2012 holiday shopping season. Imagine the return on investment those companies saw with two to three times as much engagement with users during that shopping season. It was all due to using deep learning to make better decisions around advertising campaigns.
While the science behind deep learning may be a bit difficult to understand, the message for business leaders is a simple one: Deep learning systems help you put your data to work to make better decisions, build new revenue streams, and protect your organization and your customers. To get started, you just need to be willing to put the time, money, and effort into finding and implementing the right systems and talent.
To learn more about the future of deep learning and big data, read Predictions for Big Data Analytics in 2016.