Keras: Making Deep Learning More Accessible
People often mistakenly use deep learning and machine learning as synonyms. Actually, it would be more correct to consider deep learning a subset of machine learning and about as similar to traditional machine learning as a calculator is to a computer. Deep learning relies on neural nets and can process massive quantities of data to train itself to make impressive decisions.
For instance, developers created a chess program that could defeat world champions over 20 years ago; however, two years ago, Google's Alpha Zero not only beat the former candidate for the world's best chess engine, it taught itself to play. Instead of relying upon a vast memory and training from humans, like typical machine learning software, Alpha Zero learned the basics of the game and played against itself millions' of times, thereby learning the strategies and patterns that led to wins.
How Keras Can Help Introduce the Benefits of Deep Learning to More Companies?
The benefits of a machine that can not only make complex decisions but efficiently train itself to do so could revolutionize all sorts of organizations. Previous frameworks have just been so complicated to use that a shortage of developers and hiring expenses created major obstacles. predicted that the talent shortage will continue for years in the future. While the number of job postings has doubled in the past couple of years, the number of job seekers has remained flat. In the competitive market for top talent, companies must offer deep learning engineers six figures and generous perks.
A simpler way to work with deep learning may offer a solution. Developed in Python, Keras provides an API to interface with a variety of deep learning engines. Backers include Google, Amazon, Microsoft, Apple, and many other companies. Keras also can run on many different platforms, including IoS, Android, and even within a web browser. Keras users benefit from the user-friendly and modular design that makes it relatively easy to get started with. Martin Heller, an editor at InfoWorld, says, "Keras makes deep learning about as simple as deep learning can be."
How does Onyx use Keras?
As part of Onyx’ endeavor to bring advanced analytics to our customers, both current and future, we are employing Keras as one of the underpinnings of our framework and capability. Whether it be deploying models for analysis of data at rest in lakes or in aggregating and evaluating streaming data, in conjunction with Tensorflow, Spark or Kafka, it is proving to be a powerful tool in uncovering the nuances that go unnoticed by the human analyst.
As part of the Cyber Security Analytic support we provide our customers, we utilize Keras and Tensorflow to detect malicious activity at “the edge” following a few patterns that are widely employed. In one case, specifically, we created a flow and deployed and trained a Recurrent Neural Network (RNN) to perform real-time characterization and threat detection of edge logs. The result was the detection of 42% more potential threats that had previously gone unnoticed.
For specifics on this engagement or the engineering and data science behind our efforts feel free to reach out to us.