Deep Learning Outperforming Classic ML Methods, Says Report

A report "Deep Learning: Opportunities and Best Practice" by Peltarion which is a leading AI innovator and creator of operational deep learning platform and AI knowledge Network CognitionX, organizer of the CogX festival of AI and emerging technology, illustrates that deep learning is overtaking more classic ML methods.

The study is based on the research and interviews with other AI industry players across the European region including Amazon, Google, DeepMind, and JP Morgan.

The report also depicts that, several challenges still exist like – cost, complexity, and skills which are yet to be solved to enable market growth.

The report will serve as a handbook for those who have little knowledge about deep learning and a guide for those who have more experience. It gives a better understanding of deep learning, where it's moving, and best practices from the field.

The co-founder of CognitionX, Tabitha Goldstaub stated, "At a time when so many organizations are debating the risks and rewards of using AI, I'm thrilled to see this report give some practical guidance to businesses on where deep learning can be applied, along with options on how to deliver this technology and some pointers on where the technique will go in the future. This is a good starting point for business leaders who are thinking about adopting AI."

The study odes a brief history of artificial intelligence in general and detailed comprehension of deep learning. It also incorporates the case stories of real-world applications from different verticals, such as – Pattern-recognition in healthcare diagnostics; Real-time prediction of fraudulent transactions; and Automation of complex tasks in manufacturing workflows.

For deep learning to meet such potential, it still has to overcome certain challenges. As the report quotes – "As with any large-scale IT project, deep learning projects often fail due to factors such as complexity, failure to clearly define requirements and lack of proper communication between business and technical teams."

If an organization's data is not in order, and it lacks talent in its team, these challenges get worse when tools are expensive and complicated.

Scott Penberthy, Director of Applied AI at Google and report contributor, said – "We believe there are about 10,000 people in the world who really understand DL. There are about 100,000 deep learning practitioners and a million data scientists."

Additionally, Luka Crnkovic-Friis, the Co-Founder and CEO of Peltarion, and deep learning expert believes that for deep learning to reach its full potential, it needs to be operationalized. He further added that "AI and DL will save millions of lives and improve the lives of billions. The technology will fundamentally impact health, food production, energy, business, and creativity. But if the true potential of AI and deep learning is going to be reached, it needs to be practically accessible by innovators across the world – the many, not just a few. One of the suggested routes to making deep learning more accessible is via a platform model, which simplifies and automates many tasks and provides the capability for managing the end-to-end DL workflow in one place, with an easier transition to running these models in live production environments."