Machine Learning – A Picture Is Worth A Thousand Words

[column width=”1/1″ last=”true” title=”” title_type=”single” animation=”none” implicit=”true”]

According to Wikipedia, Pointillism is a technique of painting in which small, distinct dots of color are applied in patterns to form an image. The technique relies on the ability of the eye and mind of the viewer to blend the color spots into a fuller range of tones. The practice of Pointillism is in sharp contrast to the traditional methods of blending pigments on a palette. Pointillism is analogous to the four-color CMYK printing process used by some color printers and large presses that place dots of Cyan (blue), Magenta (red), Yellow, and Key (black). Televisions and computer monitors use a similar technique to represent image colors using Red, Green, and Blue (RGB) colors.

Modern technology has given us the wonderful ability to collect data that we only dreamed about a decade ago. Three decades ago, we had to guess what our customers were thinking or doing based on limited data and a keen intellect mixed with hunches. If you could draw a picture at all, it probably looked like big polka-dots drawn using crayons. Today, we are overwhelmed with excess data. We have so much of it, it’s easy to get lost in the noise.

Let me use a medical/health example. In the not so distant past, health was measured infrequently through visits to the family doctor. Vitals such as heart rate, blood pressure, temperature, etc were recorded and compared to past data points to get a trend. If the trend was not in the desired direction, further action could be taken. The issue for doctors is getting to the heart of what cause and effects might be occurring. With wearable sensors such as watches and biometric monitoring, we can generate thousands of data points per day and millions per month! The good news is we have a much more complete picture, but the bad news is we have too much data to easily interpret at a glance.

Big data needs machine learning techniques to help us bring the picture to life and separate the true gems of knowledge from the noise. Sometimes this can be done by automatic algorithms, but to get to the real story, you need a talented person who knows what the data implicitly means, and can create custom machine learning techniques to recognize or model patterns. In the health example, this might be a scientific/medical researcher who understands the cause and effect math in the data and can create models to differentiate a simple fever or cold from the onset of cancer or heart disease.

I just read a very lengthy article from the perspective of a VC firm expressing all of the ways machine learning has manifested itself in our various ecosystems. The article suggested all of the ways that embedding machine learning into larger software platforms is adding value, and that once large corporations jump the seemingly large hurdle of understanding the implications, they pursue the path with a newfound passion.

Let me try to simplify the concept of machine learning. Machine learning at it’s core is simply a dynamic mathematical model. Imagine that you could predict an outcome based on eight indicators. The eight indicators can predict any number of potential outcomes. Let’s say that we are trying to predict ticket sales at any time of day throughout the year in all seasons. That is not going to be a linear scale, it could be anywhere on the chart. But knowing all of the combinations of the eight indicators can give us a very high confidence in our prediction. You train a model to recognize all of the ways those eight indicators equal a certain outcome. Then when you feed the model a random set of eight indicators, it will tell you which outcome to expect and also give you how confident it is in that outcome as a percentage, 100% being the most confident.

When all of the combinations of these eight data points (in this specific example) and outcomes have been used to create a machine learning model, they essentially paint a picture of all possible outcomes and allow the viewer to increase their knowledge based on the experience of the machine learning model.

It is as simple as that.

Bonus Thought: Please be aware that Deep Learning is higher up the food chain that machine learning, and Artificial Intelligence is even higher up the food chain. You could say that Deep Learning is the art of assembling multiple Machine Learning models in parallel or sequence. Artificial Intelligence makes use of multiple Machine Learning and Deep Learning models to make even higher level decisions.

[/column]