Unveiling the Effectiveness of Latent Representations in AI Systems

Published on February 16, 2023

Just like ingredients are essential to a recipe, latent representations are vital to cognitive artificial intelligence (AI) systems. In this study, we delve into the performance of different sequential clustering algorithms on the latent representations created by autoencoder and convolutional neural network (CNN) models. We also introduce a groundbreaking algorithm called Collage, which incorporates views and concepts into sequential clustering to bridge the gap with cognitive AI. By reducing memory usage and hardware operations, Collage improves energy efficiency, speed, and area performance of AI accelerators. The results reveal that while plain autoencoders produce latent representations with significant overlap between clusters, CNNs address this issue to some extent. However, CNNs introduce their own challenges within generalized cognitive pipelines.

Latent representations are a necessary component of cognitive artificial intelligence (AI) systems. Here, we investigate the performance of various sequential clustering algorithms on latent representations generated by autoencoder and convolutional neural network (CNN) models. We also introduce a new algorithm, called Collage, which brings views and concepts into sequential clustering to bridge the gap with cognitive AI. The algorithm is designed to reduce memory requirements, numbers of operations (which translate into hardware clock cycles) and thus improve energy, speed and area performance of an accelerator for running said algorithm. Results show that plain autoencoders produce latent representations which have large inter-cluster overlaps. CNNs are shown to solve this problem, however introduce their own problems in the context of generalized cognitive pipelines.

Read Full Article (External Site)

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes:

<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>