Inspired by progress in large-scale language modeling, we apply a similar approach towards building a single generalist agent beyond the realm of text outputs. The agent, which we refer to as Gato, works as a multi-modal, multi-task, multi-embodiment generalist policy. The same network with the same weights can play Atari, caption images, chat, stack blocks with a real robot arm and much more, deciding based on its context whether to output text, joint torques, button presses, or other tokens. In this report we describe the model and the data, and document the current capabilities of Gato.

Reliable and cost-effective counting of people in large indoor spaces is a significant challenge with many applications. An emerging approach is to deploy multiple fisheye cameras mounted overhead to monitor the whole space. However, due to the overlapping fields of view, person re-identificaiton (PRID) is critical for the accuracy of counting. While PRID has been thoroughly researched for traditional rectilinear cameras, few methods have been proposed for fisheye cameras and their performance is comparatively lower. To close this performance gap, we propose a multi-feature framework for fisheye PRID where we combine deep-learning, color-based and location-based features by means of novel feature fusion. We evaluate the performance of our framework for various feature combinations on FRIDA, a public fisheye PRID dataset. The results demonstrate that our multi-feature approach outperforms recent appearance-based deep-learning methods by almost 18% points and location-based methods by almost 3% points in accuracy.

Accurate activity location prediction is a crucial component of many mobility applications and is particularly required to develop personalized, sustainable transportation systems. Despite the widespread adoption of deep learning models, next location prediction models lack a comprehensive discussion and integration of mobility-related spatio-temporal contexts. Here, we utilize a multi-head self-attentional (MHSA) neural network that learns location transition patterns from historical location visits, their visit time and activity duration, as well as their surrounding land use functions, to infer an individual's next location. Specifically, we adopt point-of-interest data and latent Dirichlet allocation for representing locations' land use contexts at multiple spatial scales, generate embedding vectors of the spatio-temporal features, and learn to predict the next location with an MHSA network. Through experiments on two large-scale GNSS tracking datasets, we demonstrate that the proposed model outperforms other state-of-the-art prediction models, and reveal the contribution of various spatio-temporal contexts to the model's performance. Moreover, we find that the model trained on population data achieves higher prediction performance with fewer parameters than individual-level models due to learning from collective movement patterns. We also reveal mobility conducted in the recent past and one week before has the largest influence on the current prediction, showing that learning from a subset of the historical mobility is sufficient to obtain an accurate location prediction result. We believe that the proposed model is vital for context-aware mobility prediction. The gained insights will help to understand location prediction models and promote their implementation for mobility applications.

We address 2D floorplan reconstruction from 3D scans. Existing approaches typically employ heuristically designed multi-stage pipelines. Instead, we formulate floorplan reconstruction as a single-stage structured prediction task: find a variable-size set of polygons, which in turn are variable-length sequences of ordered vertices. To solve it we develop a novel Transformer architecture that generates polygons of multiple rooms in parallel, in a holistic manner without hand-crafted intermediate stages. The model features two-level queries for polygons and corners, and includes polygon matching to make the network end-to-end trainable. Our method achieves a new state-of-the-art for two challenging datasets, Structured3D and SceneCAD, along with significantly faster inference than previous methods. Moreover, it can readily be extended to predict additional information, i.e., semantic room types and architectural elements like doors and windows. Our code and models will be available at: https://github.com/ywyue/RoomFormer.

The well-documented presence of texture bias in modern convolutional neural networks has led to a plethora of algorithms that promote an emphasis on shape cues, often to support generalization to new domains. Yet, common datasets, benchmarks and general model selection strategies are missing, and there is no agreed, rigorous evaluation protocol. In this paper, we investigate difficulties and limitations when training networks with reduced texture bias. In particular, we also show that proper evaluation and meaningful comparisons between methods are not trivial. We introduce BiasBed, a testbed for texture- and style-biased training, including multiple datasets and a range of existing algorithms. It comes with an extensive evaluation protocol that includes rigorous hypothesis testing to gauge the significance of the results, despite the considerable training instability of some style bias methods. Our extensive experiments, shed new light on the need for careful, statistically founded evaluation protocols for style bias (and beyond). E.g., we find that some algorithms proposed in the literature do not significantly mitigate the impact of style bias at all. With the release of BiasBed, we hope to foster a common understanding of consistent and meaningful comparisons, and consequently faster progress towards learning methods free of texture bias. Code is available at https://github.com/D1noFuzi/BiasBed

Tasks critical to enterprise profitability, such as customer churn prediction, fraudulent account detection or customer lifetime value estimation, are often tackled by models trained on features engineered from customer data in tabular format. Application-specific feature engineering adds development, operationalization and maintenance costs over time. Recent advances in representation learning present an opportunity to simplify and generalize feature engineering across applications. When applying these advancements to tabular data researchers deal with data heterogeneity, variations in customer engagement history or the sheer volume of enterprise datasets. In this paper, we propose a novel approach to encode tabular data containing customer transactions, purchase history and other interactions into a generic representation of a customer's association with the business. We then evaluate these embeddings as features to train multiple models spanning a variety of applications. CASPR, Customer Activity Sequence-based Prediction and Representation, applies Transformer architecture to encode activity sequences to improve model performance and avoid bespoke feature engineering across applications. Our experiments at scale validate CASPR for both small and large enterprise applications.

Fine-grained population maps are needed in several domains, like urban planning, environmental monitoring, public health, and humanitarian operations. Unfortunately, in many countries only aggregate census counts over large spatial units are collected, moreover, these are not always up-to-date. We present POMELO, a deep learning model that employs coarse census counts and open geodata to estimate fine-grained population maps with 100m ground sampling distance. Moreover, the model can also estimate population numbers when no census counts at all are available, by generalizing across countries. In a series of experiments for several countries in sub-Saharan Africa, the maps produced with POMELOare in good agreement with the most detailed available reference counts: disaggregation of coarse census counts reaches R2 values of 85-89%; unconstrained prediction in the absence of any counts reaches 48-69%.

我们提出了GRASP提案网络（GP-NET），这是一种卷积神经网络模型，可以为移动操纵器生成6-DOF GRASP。为了训练GP-NET，我们合成生成一个包含深度图像和地面真相掌握信息的数据集，以供超过1400个对象。在现实世界实验中，我们使用egad！掌握基准测试，以评估两种常用算法的GP-NET，即体积抓地力网络（VGN）和在PAL TIAGO移动操纵器上进行的GRASP抓取网络（VGN）和GRASP姿势检测包（GPD）。GP-NET的掌握率为82.2％，而VGN为57.8％，GPD的成功率为63.3％。与机器人握把中最新的方法相反，GP-NET可以在不限制工作空间的情况下使用移动操纵器抓住对象，用于抓住对象，需要桌子进行分割或需要高端GPU。为了鼓励使用GP-NET，我们在https://aucoroboticsmu.github.io/gp-net/上提供ROS包以及我们的代码和预培训模型。

积极和未标记的学习是一个重要的问题，在许多应用中自然出现。几乎所有现有方法的显着局限性在于假设倾向得分函数是恒定的（疤痕假设），这在许多实际情况下都是不现实的。避免这种假设，我们将参数方法考虑到后验概率和倾向得分功能的关节估计问题。我们表明，在轻度假设下，当两个函数具有相同的参数形式（例如，具有不同参数的逻辑）时，相应的参数是可识别的。在此激励的情况下，我们提出了两种估计方法：关节最大似然法和第二种方法基于两种Fisher一致表达式的交替实现。我们的实验结果表明，所提出的方法比基于预期最大化方案的现有方法可比性或更好。

随时间变化数据的因果发现（CD）在神经科学，医学和机器学习中很重要。 CD的技术包括通常没有偏见但昂贵的随机实验。它还包括诸如回归，匹配和Granger因果关系之类的算法，这些算法仅在人类设计师做出的强烈假设下正确。但是，正如我们在机器学习的其他领域发现的那样，人类通常不太正确，通常比数据驱动的方法表现出色。在这里，我们测试是否可以以数据驱动的方式改善因果发现。我们采用了一个具有大量因果成分（晶体管），MOS 6502处理器和元学习的系统，该系统为神经网络代表的因果发现程序。我们发现，该程序的表现远远超过了人为设计的因果发现程序，例如相互信息和Granger因果关系。我们认为，因果关系领域应在可能的情况下考虑一种有监督的方法，其中从具有已知因果关系的大型数据集中学习了CD程序，而不是由人类专家设计。我们的发现有望在神经和医学数据以及更广泛的机器学习社区中采用新的CD方法。