Machine Learning and Deep Learning
Every industry – from precision medicine to precision agriculture, from personalized retail to personalized banking – is being impacted by Machine Learning.
Machine learning (ML) and deep learning have allowed for completely new possibilities in the realm of digital experiences, targeting and predictions. Companies in every industry are collecting more and more data. It’s now possible to train analytical models that even a few years ago would have been impossible. Business has taken notice of their data sets’ power, leading them to develop completely new products and initiatives.
What is Machine Learning(ML)? ML is a part of the broader fields of Computer Science and Statistics. The goal of ML is to enable computers to learn from their experience in certain tasks. ML also enables the machine to improve performance as their experience grows. A self-driving car, for example, learns from being initially driven around by a human driver; further, as it drives itself, it reinforces its own learning and gets better with experience. In finance, one can view ML as an attempt at uncovering relationships between variables, where given historical patterns (input and output), the machine forecasts outcomes out of sample.
ML can also be seen as a model-independent (or statistical or data-driven) way for recognizing patterns in large data sets. ML techniques include:
- Supervised Learning (methods such as regressions and classifications)
- Unsupervised Learning (factor analyses and regime identification)
- Deep and Reinforced Learning. Deep learning is based on neural network algorithms, and is used in processing unstructured data (e.g. images, voice, sentiment, etc.) and pattern recognition in structured data.
What has changed in the past few years? Machine learning is moving from the fringe (doing data science in silos with dedicated PhDs and data scientists with a variety of niche tools (R, scikit-learn, Tensorflow, etc.)) to the core, supported by Machine Learning (ML) as a service. ML as a service model means business teams can move from a focus on custom engineering to model building and deployment by leveraging a variety of algorithms, common APIs, computing platforms and data services.
In many digital business models which function as a two-sided marketplace – AirBnB, Uber, Lyft, Amazon.com etc. — where you have buyers (or riders) on one side and sellers (or drivers) on the other side, an efficient marketplace comes from dynamic matching, predictions and attribution. Three-sided marketplaces – UberEats (consumers, delivery vehicles/drivers, restaurant – meal preparation) take the complexity to another level. Core platforms are becoming increasingly smart and dynamic learning applications that benefit from knowing user behavior via click-thrus, selections and likes and from knowing people’s historical behavior.
A good example of this trend is Uber’s Michelangelo, an internal ML-as-a-service platform that makes scaling AI possible for a variety of prediction use cases. Michelangelo enables internal teams to seamlessly build, deploy, and operate machine learning solutions at Uber’s scale. It is designed to cover the end-to-end ML workflow: manage data, train, evaluate, and deploy models, make predictions, and monitor predictions. The system also supports traditional ML models, time series forecasting, and deep learning.
According to a Uber Engineering blog post, Michelangelo has been serving production use cases at Uber since 2016 and has become the de-facto system for ML for engineers and data scientists, with dozens of teams building and deploying models. It is deployed across several Uber data centers, leverages specialized hardware like Nvidia GPUs, and serves predictions for the heavily used online services at the company. ML is used to enable an efficient ride-sharing marketplace, identify suspicious or fraudulent accounts, and suggest optimal pickup and dropoff points.
Uber is not unique. Every firm looking for the next competitive edge in digital is racing to support the six-step ML workflows: (1) Manage data; (2) Train models; (3) Evaluate models; (4) Deploy models; (5) Make predictions and (6) Monitor predictions.
According to Mariya Yao in Medium: breakthroughs in deep learning have driven major advances in machine perception. Computers can now reliably detect and classify objects in images and video, transcribe and translate speech as well as humans, and even generate art, music, and movie soundtracks.
We have clearly entered the era of ML and AI in digital. The What and Why are becoming clear. The race for creating and exploiting the next killer app is underway.
- Machine learning (ML) is functionality that helps software applications perform a task without explicit programming or rules. Traditionally considered a subcategory of AI, ML involves statistical techniques, such as deep learning (aka neural networks), that are inspired by neuroscience theories about how the human brain processes information. Methods of Machine Learning include: