Agrim Gupta

Title: Agrim Gupta and Temporal Sequence Generation

Introduction

Agrim Gupta is an inventor based in Stanford, California, and works at Google Inc. His patent work addresses the generation of temporal sequences using diffusion transformer neural networks.

Latest Patents

Gupta is associated with 1 patent, titled “Generating Temporal Sequences Using Diffusion Transformer Neural Networks.” The invention describes methods, systems, and computer programs for generating an output sequence of data elements conditioned on an input. The input includes noise-based latent representations for multiple sequence segments. A latent denoising neural network updates these representations over multiple iterations, using memory vectors that capture hidden states from preceding segments. The resulting latent representations are then processed to generate the output sequence.

Career Highlights

Gupta’s listed invention focuses on applying neural-network denoising and information from preceding segments to temporal sequence generation. His work is associated with Google Inc.

Collaborations

The listed collaborators include Sihyun Yu and Meera Satya Hahn.

Conclusion

Agrim Gupta’s patent describes a neural-network approach to generating temporal data sequences, combining iterative latent updates with information carried forward from earlier segments.