Three-Dimensional Digital Template Atlas of the Macaque Brain.

We present a new 3D template atlas of the anatomical subdivisions of the macaque brain, which is based on and aligned to the magnetic resonance imaging (MRI) data set and histological sections of the Saleem and Logothetis atlas. We describe the creation and validation of the atlas that, when registered with macaque structural or functional MRI scans, provides a straightforward means to estimate the boundaries between architectonic areas, either in a 3D volume with different planes of sections, or on an inflated brain surface (cortical flat map). As such, this new template atlas is intended for use as a reference standard for macaque brain research. Atlases and templates are available as both volumes and surfaces in standard NIFTI and GIFTI formats.

Mastering the game of Stratego with model-free multiagent reinforcement learning.

We introduce DeepNash, an autonomous agent that plays the imperfect information game Stratego at a human expert level. Stratego is one of the few iconic board games that artificial intelligence (AI) has not yet mastered. It is a game characterized by a twin challenge: It requires long-term strategic thinking as in chess, but it also requires dealing with imperfect information as in poker. The technique underpinning DeepNash uses a game-theoretic, model-free deep reinforcement learning method, without search, that learns to master Stratego through self-play from scratch. DeepNash beat existing state-of-the-art AI methods in Stratego and achieved a year-to-date (2022) and all-time top-three ranking on the Gravon games platform, competing with human expert players.

Navigating the landscape of multiplayer games.

Multiplayer games have long been used as testbeds in artificial intelligence research, aptly referred to as the Drosophila of artificial intelligence. Traditionally, researchers have focused on using well-known games to build strong agents. This progress, however, can be better informed by characterizing games and their topological landscape. Tackling this latter question can facilitate understanding of agents and help determine what game an agent should target next as part of its training. Here, we show how network measures applied to response graphs of large-scale games enable the creation of a landscape of games, quantifying relationships between games of varying sizes and characteristics. We illustrate our findings in domains ranging from canonical games to complex empirical games capturing the performance of trained agents pitted against one another. Our results culminate in a demonstration leveraging this information to generate new and interesting games, including mixtures of empirical games synthesized from real world games.

A new fast accurate nonlinear medical image registration program including surface preserving regularization.

Recently inexpensive graphical processing units (GPUs) have become established as a viable alternative to traditional CPUs for many medical image processing applications. GPUs offer the potential of very significant improvements in performance at low cost and with low power consumption. One way in which GPU programs differ from traditional CPU programs is that increasingly elaborate calculations per voxel may not impact of the overall processing time because memory accesses can dominate execution time. This paper presents a new GPU based elastic image registration program named Ezys. The Ezys image registration algorithm belongs to the wide class of diffeomorphic demons but uses surface preserving image smoothing and regularization filters designed for a GPU that would be computationally expensive on a CPU. We describe the methods used in Ezys and present results from two important neuroscience applications. Firstly inter-subject registration for transfer of anatomical labels and secondly longitudinal intra-subject registration to quantify atrophy in individual subjects. Both experiments showed that Ezys registration compares favorably with other popular elastic image registration programs. We believe Ezys is a useful tool for neuroscience and other applications, and also demonstrates the value of developing of novel image processing filters specifically designed for GPUs.