English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Unsupervised multi-object representation learning depends on inductive biasesto guide the discovery of object-centric representations that generalize.However, we observe that methods for learning these representations are eitherimpractical due to long training times and large memory consumption or foregokey inductive biases. In this work, we introduce EfficientMORL, an efficientframework for the unsupervised learning of object-centric representations. Weshow that optimization challenges caused by requiring both symmetry anddisentanglement can in fact be addressed by high-cost iterative amortizedinference by designing the framework to minimize its dependence on it. We takea two-stage approach to inference: first, a hierarchical variationalautoencoder extracts symmetric and disentangled representations throughbottom-up inference, and second, a lightweight network refines therepresentations with top-down feedback. The number of refinement steps takenduring training is reduced following a curriculum, so that at test time withzero steps the model achieves 99.1% of the refined decomposition performance.We demonstrate strong object decomposition and disentanglement on the standardmulti-object benchmark while achieving nearly an order of magnitude fastertraining and test time inference over the previous state-of-the-art model.

Efficient Iterative Amortized Inference for Learning Symmetric and Disentangled Multi-Object Representations