Staircase model, Transformer cores are stacked diagonally, so each step sees one new input chunk.

As a fundamental problem in computer vision, point cloud registration aims to seek the optimal transformation for aligning a pair of point clouds. In most existing methods, the information flows are usually forward transferring, thus lacking the guidance from high-level information to low-level information. Besides, excessive high-level information may be overly redundant, and directly using it may conflict with the original low-level information. In this paper, we propose a novel Iterative Feedback Network (IFNet) for unsupervised point cloud registration, in which the representation of low-level features is efficiently enriched by rerouting subsequent high-level features. Specifically, our IFNet is built upon a series of Feedback Registration Block (FRB) modules, with each module responsible for generating the feedforward rigid transformation and feedback high-level features. These FRB modules are cascaded and recurrently unfolded over time. Further, the Feedback Transformer is designed to efficiently select relevant information from feedback high-level features, which is utilized to refine the low-level features. What's more, we incorporate a geometry-awareness descriptor to empower the network for making full use of most geometric information, which leads to more precise registration results. Extensive experiments on various benchmark datasets demonstrate the superior registration performance of our IFNet.

GPT-3 has created a lot of buzz since its release a few months ago. The system can generate (almost) plausible conversations with the likes of Nietzsche, write op eds for The Guardian and was even used successfully to post undercover comments on Reddit for a week. But even with GPT-3, AI is still stuck in Uncanny Valley. GPT-3 output feels like it was written by a human at first glance, but it isn't quite. On closer inspection, it lacks substance and coherence.

Transformers are feedforward networks that can process input tokens in parallel. While this parallelization makes them computationally efficient, it restricts the model from fully exploiting the sequential nature of the input - the representation at a given layer can only access representations from lower layers, rather than the higher level representations already built in previous time steps. In this work, we propose the Feedback Transformer architecture that exposes all previous representations to all future representations, meaning the lowest representation of the current timestep is formed from the highest-level abstract representation of the past. We demonstrate on a variety of benchmarks in language modeling, neural machine translation, summarization, and reinforcement learning that the increased representation capacity can improve over Transformer baselines.