Search

As a foundational component of cognitive intelligence, theory of mind (ToM) can make AI more closely resemble human thought processes, thereby enhancing their interaction and collaboration with human. In particular, it can significantly improve a model's comprehension of videos in complex scenes. However, current video question answer (VideoQA) datasets focus on studying causal reasoning within events few of them genuinely incorporating human ToM. Consequently, there is a lack of development in ToM reasoning tasks within the area of VideoQA. This paper presents BDIQA, the first benchmark to explore the cognitive reasoning capabilities of VideoQA models in the context of ToM. BDIQA is inspired by the cognitive development of children's ToM and addresses the current deficiencies in machine ToM within datasets and tasks. Specifically, it offers tasks at two difficulty levels, assessing Belief, Desire and Intention (BDI) reasoning in both simple and complex scenarios. We conduct evaluations on several mainstream methods of VideoQA and diagnose their capabilities with zero shot, few shot and supervised learning. We find that the performance of pre-trained models on cognitive reasoning tasks remains unsatisfactory. To counter this challenge, we undertake thorough analysis and experimentation, ultimately presenting two guidelines to enhance cognitive reasoning derived from ablation analysis.

Theory of Mind (ToM) is the ability to attribute mental states to others, the basis of human cognition. At present, there has been growing interest in the AI with cognitive abilities, for example in healthcare and the motoring industry. Beliefs, desires, and intentions are the early abilities of infants and the foundation of human cognitive ability, as well as for machine with ToM. In this paper, we review recent progress in machine ToM on beliefs, desires, and intentions. And we shall introduce the experiments, datasets and methods of machine ToM on these three aspects, summarize the development of different tasks and datasets in recent years, and compare well-behaved models in aspects of advantages, limitations and applicable conditions, hoping that this study can guide researchers to quickly keep up with latest trend in this field. Unlike other domains with a specific task and resolution framework, machine ToM lacks a unified instruction and a series of standard evaluation tasks, which make it difficult to formally compare the proposed models. We argue that, one method to address this difficulty is now to present a standard assessment criteria and dataset, better a large-scale dataset covered multiple aspects of ToM., Comment: arXiv admin note: text overlap with arXiv:2102.11938 by other authors

This paper mainly studies the gradient-based Jacobi-type algorithms to maximize two classes of homogeneous polynomials with orthogonality constraints, and establish their convergence properties. For the first class of homogeneous polynomials subject to a constraint on a Stiefel manifold, we reformulate it as an optimization problem on a unitary group, which makes it possible to apply the gradient-based Jacobi-type (Jacobi-G) algorithm. Then, if the subproblem can always be represented as a quadratic form, we establish the global convergence of Jacobi-G under any one of three conditions. The convergence result for the first condition is an easy extension of the result in [Usevich et al. SIOPT 2020], while other two conditions are new ones. This algorithm and the convergence properties apply to the well-known joint approximate symmetric tensor diagonalization. For the second class of homogeneous polynomials subject to constraints on the product of Stiefel manifolds, we reformulate it as an optimization problem on the product of unitary groups, and then develop a new gradient-based multi-block Jacobi-type (Jacobi-MG) algorithm to solve it. We establish the global convergence of Jacobi-MG under any one of the above three conditions, if the subproblem can always be represented as a quadratic form. This algorithm and the convergence properties are suitable to the well-known joint approximate tensor diagonalization. As the proximal variants of Jacobi-G and Jacobi-MG, we also propose the Jacobi-GP and Jacobi-MGP algorithms, and establish their global convergence without any further condition. Some numerical results are provided indicating the efficiency of the proposed algorithms., Comment: 23 pages, 3 figures

Cortical actin networks are highly dynamic and play critical roles in shaping the mechanical properties of cells. The actin cytoskeleton undergoes significant reorganization over the course of the cell cycle, when cortical actin transitions between open patched meshworks, homogeneous distributions, and aligned bundles. Several types of myosin motor proteins, characterized by different kinetic parameters, have been involved in this reorganization of actin filaments. Given the limitations in studying the interactions of actin with myosin in vivo, we propose stochastic agent-based model simulations and develop a set of data analysis measures to assess how myosin motor proteins mediate various actin organizations. In particular, we identify individual motor parameters, such as motor binding rate and step size, that generate actin networks with different levels of contractility and different patterns of myosin motor localization. In simulations where two motor populations with distinct kinetic parameters interact with the same actin network, we find that motors may act in a complementary way, by tuning the actin network organization, or in an antagonistic way, where one motor emerges as dominant. This modeling and data analysis framework also uncovers parameter regimes where spatial segregation between motor populations is achieved. By allowing for changes in kinetic rates during the actin-myosin dynamic simulations, our work suggests that certain actin-myosin organizations may require additional regulation beyond mediation by motor proteins in order to reconfigure the cytoskeleton network on experimentally-observed timescales., Comment: 24 pages, 11 figures

Simulating soft matter systems such as the cytoskeleton can enable deep understanding of experimentally observed phenomena. One challenge of modeling such systems is realistic description of the steric repulsion between nearby polymers. Previous models of the polymeric excluded volume interaction have the deficit of being non-analytic, being computationally expensive, or allowing polymers to erroneously cross each other. A recent solution to these issues, implemented in the MEDYAN simulation platform, uses analytical expressions obtained from integrating an interaction kernel along the lengths of two polymer segments to describe their repulsion. Here, we extend this model by re-deriving it for lower-dimensional geometrical configurations, deriving similar expressions using a steeper interaction kernel, comparing it to other commonly used potentials, and showing how to parameterize these models. We also generalize this new integrated style of potential by introducing a segmental Lennard-Jones potential, which enables modelling both attractive and repulsive interactions in semi-flexible polymer networks. These results can be further generalized to facilitate the development of effective interaction potentials for other finite elements in simulations of soft-matter systems., Comment: 18 pages, 6 figures

In this paper, a new alternating direction trust region method based on conic model is used to solve unconstrained optimization problems. By use of the alternating direction method, the new conic model trust region subproblem is solved by two steps in two orthogonal directions. This new idea overcomes the shortcomings of conic model subproblem which is difficult to solve. Then the global convergence of the method under some reasonable conditions is established. Numerical experiment shows that this method may be better than the dogleg method to solve the subproblem, especially for large-scale problems., Comment: 18 pages,3 tables