What is deep learning?

One of the crucial recent developments in machine learning is deep learning. This is a combined term for algorithms that both train and use very deep neural networks.

The impact of deep learning

The impact of deep learning on the performance of machine-learning algorithms is enormous. Tasks like image recognition, sound recognition and text classification now have unprecedented performance levels because of these techniques. Therefore, you can find deep learning applied in just about any AI application, from speech recognition in your cell phone to self-driving cars, computers that can beat professional Go players, and even search engines.

Neural networks at work

The main idea behind these neural networks is that by adding a lot of layers to it, the algorithm can find complex features from data on its own. You can, for example, input raw images to the algorithm and the network will identify representative features that describe the images it sees. This saves a lot of time and effort in feature engineering. As a consequence, performance is a lot better than previously possible.

Prof. Max Welling explains the mechanism and power of deep learning.

Apply deep learning with us/

Leveraging this type of technology to its fullest potential is our specialty. With an ever-expanding amount of research available, a multitude of different libraries and implementations, and new applications constantly being developed, deep learning requires highly specific expertise kept up to date nonstop. And that’s exactly what we do at Scyfer.

Developing and deploying deep learning models

We’ve developed our own in-house framework to rapidly develop and deploy deep-learning models. We’ve built it around Theano, with custom-built features based on the latest developments in the field to train neural networks for peak performance.

Our deep-learning platform optimizes for speed, with a comprehensive dashboard for tracking experiments. It reduces complexity and includes extensive unit-testing and many other features that enable us to rapidly set up new experiments.

Do you have any questions, or an idea you want to discuss? Then contact us.

Scyfer logo