Through their collaboration, DeepLearning.AI and Landing AI have launched the Data-Centric AI Competition: an artificial intelligence challenge. Participants will be asked to improve a dataset according to a fixed pattern with the goal of improving the performance of machine learning models. Andrew Ng himself, a world-renowned expert in machine learning and robotics, has proposed that developers, coders and AI specialists participate in this challenge.

A machine learning competition that wants to break the traditional codes

The challenge proposed by Landing AI and DeepLearning.AI wants to “reverse the traditional format” of machine learning competitions where participants are asked to create a high-performance model from a fixed data set. The organisers state:

“machine learning has matured to the point where high-performance model architectures are widely available, while dataset engineering approaches have lagged behind.”

Participants will be asked to improve a dataset based on a model already proposed. A dataset will be provided to developers: they will have to enrich it by applying data-centric techniques such as correcting incorrect labels, applying data augmentation or adding examples representing extreme or rare cases.

How to participate?

As of June 17, applicants can submit their modified dataset for evaluation no later than September 4, 2021, the anniversary of the birth date of John McCarthy who coined the terminology “artificial intelligence” and was the main pioneer of the field.

Three winners will be chosen in two categories: “best overall performance” and “most innovative data set”. The winners will be invited to meet Andrew Ng at a private event to discuss and share ideas on how to develop initiatives to improve databases and data in general. These ideas will be promoted on the various channels of Landing AI and DeepLearning.AI

To participate, go here.

Translated from Andrew Ng lance le Data-Centric AI Competition, un défi centré sur l’amélioration de jeux de données