This Is What I Am Learning – Generative Adversarial Networks (GANs)

As an entrepreneur working at the interface of electronics and big data with the analytics, competitive capability is about finding how to create better models. To do that, you work to be at the state of the art. In the electronics world, I do believe that Generative Adversarial Networks (GANs) will take over and extend the capabilities of GPU (graphics processing units) in coming years. Implementing GANs enablers in circuits and boards will be very catalytic for those that process and manipulate data, at scale, within the whole constructs of AI.

Generative adversarial networks (GANs) are a class of artificial intelligence algorithms used in unsupervised machine learning, implemented by a system of two neural networks contesting with each other in a zero-sum game framework. This technique can generate photographs that look at least superficially authentic to human observers, having many realistic characteristics (though in tests people can tell real from generated in many cases). It was invented by Ian Goodfellow.. (Wikipedia)

Generative Adversarial Networks (GANs) has transformed deep learning by accurately modeling real world data better than any model developed before.

In Fasmicro Group, we are working to improve models used in our products, already anchored on neuromorphics which emulate human biology to make better systems. I am learning GANs to see how we can extend our capabilities as we serve clients through our products. GANs is going to be huge, and we want to have deep understanding of the applications as it scales.

Here, we have a history of tearing things apart, but in this specific one, it seems we have to build something. As Intel FPGA partner in Africa, we are exploring how to implement GANs in hardware to examine how it could improve some of the things we are doing.

Register for Tekedia Mini-MBA (Feb 10 - May 3, 2025), and join Prof Ndubuisi Ekekwe and our global faculty; click here.