United States

Vision-Language (VL) models with the Two-Tower architecture have dominated visual-language representation learning in recent years. Current VL models either use lightweight uni-modal encoders and learn to extract, align and fuse both modalities simultaneously in a deep cross-modal encoder, or feed the last-layer uni-modal representations from the deep pre-trained uni-modal encoders into the top cross-modal encoder. Both approaches potentially restrict vision-language representation learning and limit model performance. In this paper, we propose BridgeTower, which introduces multiple bridge layers that build a connection between the top layers of uni-modal encoders and each layer of the cross-modal encoder. This enables effective bottom-up cross-modal alignment and fusion between visual and textual representations of different semantic levels of pre-trained uni-modal encoders in the cross-modal encoder. Pre-trained with only 4M images, BridgeTower achieves state-of-the-art performance on various downstream vision-language tasks. In particular, on the VQAv2 test-std set, BridgeTower achieves an accuracy of 78.73%, outperforming the previous state-of-the-art model METER by 1.09% with the same pre-training data and almost negligible additional parameters and computational costs. Notably, when further scaling the model, BridgeTower achieves an accuracy of 81.15%, surpassing models that are pre-trained on orders-of-magnitude larger datasets. Code and checkpoints are available at https://github.com/microsoft/BridgeTower.

AAAI 2023

BridgeTower: Building Bridges Between Encoders in Vision-Language Representation Learning

multi-layer feature utilization

vision-language representation learning

multimodal learning

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-23 is the Thirty-Seventh AAAI Conference on Artificial Intelligence. The theme of this conference is to create collaborative bridges within and beyond AI. Like previous AAAI conferences, AAAI-23 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and two new activities: a Bridge Program and a Lab Program. Many of these activities are tailored to the theme of bridges and all are selected according to the highest standards, with additional programs for students and young researchers. 
AAAI is providing you with a conference planner, which you can use to help organize your itinerary of activities. This includes talks to attend in person, talks to attend remotely, breaks with colleagues and your site seeing activities. To access this conference planner, please go to [https://aaai-2023.takemobi.io](https://aaai-2023.takemobi.io).

In order to access this site, you need to register. If you haven't already, please register [here](https://aaai.org/Conferences/AAAI-23/registration/).


The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines.

poster

The computation of short paths in graphs with arc lengths is a pillar of graph algorithmics and network science. In a more diverse world, however, not every short path is equally valuable. For the setting where each vertex is assigned to a group (color), we provide a framework to model multiple natural fairness aspects. We seek to find short paths in which the number of occurrences of each color is within some given lower and upper bounds. Among other results, we prove the introduced problems to be computationally intractable (NP-hard and parameterized hard with respect to the number of colors) even in very restricted settings (such as each color should appear with exactly the same frequency), while also presenting an encouraging algorithmic result ("fixed-parameter tractability") related to the length of the sought solution path for the general problem.


Fair Short Paths in Vertex-Colored Graphs

Partial label learning (PLL) aims to learn from inexact data annotations where each training example is associated with a coarse candidate label set. Due to its practicability, many PLL algorithms have been proposed in recent literature. Most prior PLL works attempt to identify the ground-truth labels from candidate sets and the classifier is trained afterward by fitting the features of examples and their exact ground-truth labels. From a different perspective, we propose to enrich the feature space and raise the question ``Can label-specific features help PLL?'' rather than learning from examples with identical features for all classes. Despite its benefits, previous label-specific feature approaches rely on ground-truth labels to split positive and negative examples of each class and then conduct clustering analysis, which is not directly applicable in PLL. To remedy this problem, we propose an uncertainty-aware confidence region to accommodate false positive labels. We first employ graph-based label enhancement to yield smooth pseudo-labels and facilitate the confidence region split. After acquiring label-specific features, a family of binary classifiers is induced. Extensive experiments on both synthesized and real-world datasets are conducted and the results show that our method consistently outperforms eight baselines.

Can Label-Specific Feature Help Partial-Label Learning?

	Videos such as movies or TV episodes usually need to divide the long storyline into cohesive units, i.e., scenes, to facilitate the understanding of video semantics. The key challenge lies in finding the boundaries of scenes by comprehensively considering the complex temporal structure and semantic information. To this end, we introduce a novel Context-Aware Transformer (CAT) with a self-supervised learning framework to learn high-quality shot representations, for generating well-bounded scenes. More specifically, we design the CAT with local-global self-attentions, which can effectively consider both the long-term and short-term context to improve the shot encoding. For training the CAT, we adopt the self-supervised learning schema. Firstly, we leverage shot-to-scene level pretext tasks to facilitate the pre-training with pseudo boundary, which guides CAT to learn the discriminative shot representations that maximize intra-scene similarity and inter-scene discrimination in an unsupervised manner. Then, we transfer contextual representations for fine-tuning the CAT with supervised data, which encourages CAT to accurately detect the boundary for scene segmentation. As a result, CAT is able to learn the context-aware shot representations and provides global guidance for scene segmentation. Our empirical analyses show that CAT can achieve state-of-the-art performance when conducting the scene segmentation task on the MovieNet~\cite{RaoXXXHZL20} dataset, e.g., offering 2.15 improvements on AP.

Towards Global Video Scene Segmentation with Context-Aware Transformer

Recent years have seen a surge in research on dynamic graph representation learning, which aims to model temporal graphs that are dynamic and evolving constantly over time. However, current work typically models graph dynamics with recurrent neural networks (RNNs), making them suffer seriously from computation and memory overheads on large temporal graphs. So far, scalability of dynamic graph representation learning on large temporal graphs remains one of the major challenges. In this paper, we present a scalable framework, namely SpikeNet, to efficiently capture the temporal and structural patterns of temporal graphs. We explore a new direction in that we can capture the evolving dynamics of temporal graphs with spiking neural networks (SNNs) instead of RNNs. As a low-power alternative to RNNs, SNNs explicitly model graph dynamics as spike trains of neuron populations and enable spike-based propagation in an efficient way. Experiments on three large real-world temporal graph datasets demonstrate that SpikeNet outperforms strong baselines on the temporal node classification task with lower computational costs. Particularly, SpikeNet generalizes to a large temporal graph (2.7M nodes and 13.9M edges) with significantly fewer parameters and computation overheads. 

Scaling Up Dynamic Graph Representation Learning via Spiking Neural Networks

Scribble supervised semantic segmentation has achieved great advances in pseudo label exploitation, yet suffers insufficient label exploration for the mass of unannotated regions.
In this work, we propose a novel exploratory inference learning (EIL) framework, which facilitates efficient probe on unlabeled pixels and promotes selecting confident candidates for boosting the evolved segmentation. 
The exploration of unannotated regions is formulated as an iterative decision making process, where a policy searcher learns to infer in the unknown space and the reward to the exploratory policy is based on a contrastive measurement of candidates. 
In particular, we devise the contrastive reward with the inter-class attraction and the intra-class repulsion in the feature space w.r.t the pseudo labels. 
The unlabeled exploration and the labeled exploitation are jointly balanced to improve the segmentation, and framed in a close-looping end-to-end network.
Comprehensive evaluations on the benchmark datasets (including PASCAL VOC 2012~\cite{everingham2010pascal} and PASCAL Context~\cite{hariharan2011semantic}) demonstrate the superiority of our proposed EIL when compared with other state-of-the-art methods for the scribble-supervised semantic segmentation problem.

Exploratory Inference Learning for Scribble Supervised Semantic Segmentation

Multi-document summarization (MDS) aims to generate a summary for a number of related documents. We propose HGSUM — an MDS model that extends an encoder-decoder architecture to incorporate a heterogeneous graph to repre-
sent different semantic units (e.g., words and sentences) of the documents. This contrasts with existing MDS models which do not consider different edge types of graphs and as such do not capture the diversity of relationships in the documents.
To preserve only key information and relationships of the documents in the heterogeneous graph, HGSUM uses graph pooling to compress the input graph. And to guide HGSUM to learn the compression, we introduce an additional objective that maximizes the similarity between the compressed graph and the graph constructed from the ground-truth summary during training. HGSUM is trained end-to-end with graph similarity and standard cross-entropy objectives. Experimental results over MULTI-NEWS, WCEP-100, and ARXIV show that HGSUM outperforms state-of-the-art MDS models. The code for our model and experiments is available at: https://github.com/oaimli/HGSum.

Compressed Heterogeneous Graph for Abstractive Multi-document Summarization

Although remarkable progress has been made in recent years, current multi-exposure image fusion (MEF) research is still bounded by the lack of real ground truth, objective evaluation function, and robust fusion strategy. In this paper, we study the MEF problem from a new perspective. We don’t utilize any synthesized ground truth, design any loss function, or develop any fusion strategy. Our proposed method EMEF takes advantage of the wisdom of multiple imperfect MEF contributors including both conventional and deep learning-based methods. Specifically, EMEF consists of two main stages: pre-train an imitator network and tune the imitator in the runtime. In the first stage, we make a unified network imitate different MEF targets in a style modulation way. In the second stage, we tune the imitator network by optimizing the style code, in order to find an optimal fusion result for each input pair. In the experiment, we construct EMEF from four stateof-the-art MEF methods and then make comparisons with the individuals and several other competitive methods on the latest released MEF benchmark dataset. The promising experimental results provide evidence for our ensemble approach to “get the best of all worlds”.

EMEF: Ensemble Multi-Exposure Image Fusion

Altruistic punishment (or punishment) has been extensively shown as an important mechanism for promoting cooperation in human societies. In AI, the emergence of punishment has received much recent interest. In this paper, we contribute with a novel evolutionary game theoretic model to study the impacts of environmental feedback. Whereas a population of agents plays public goods games, there exists a third-party population whose payoffs depend not only on whether to punish or not, but also on the state of the environment (e.g., how cooperative the agents in a social dilemma are). Focusing on one-shot public goods games, we show that environmental feedback, by itself, can lead to the emergence of punishment. We analyze the co-evolution of punishment and cooperation, and derive conditions for their co-presence, co-dominance and co-extinction. Moreover, we show that the system can exhibit bistability as well as cyclic dynamics. Our findings provide a new explanation for the emergence of punishment.  On the other hand, our results also alert the need for careful design of implementing punishment in multi-agent systems, as the resulting evolutionary dynamics can be somewhat complex.

Emergence of Punishment in Social Dilemma with Environmental Feedback

Diffusion models have emerged as a powerful generative method for synthesizing high-quality and diverse set of images. In this paper, we propose a video generation method based on diffusion models, where the effects of motion are modeled in an implicit condition manner, i.e. one can sample plausible video motions according to the latent feature of frames. We improve the quality of the generated videos by proposing multiple strategies such as sampling space truncation, robustness penalty, and positional group normalization. Various experiments are conducted on datasets consisting of videos with different resolutions and different number of frames. Results show that the proposed method outperforms the state-of-the-art generative adversarial network-based methods by a significant margin in terms of FVD scores as well as perceptible visual quality. 

VIDM: Video Implicit Diffusion Models

Answer Set Programming (ASP) is a problem modeling and solving framework for several problems in KR with growing industrial applications. Also for studies of computational complexity and deeper insights into the hardness and its sources, ASP has been attracting researchers for many years. These studies resulted in fruitful characterizations in terms of complexity classes, fine-grained insights in form of dichotomy-style results, as well as detailed parameterized complexity landscapes. Recently, this lead to a novel result establishing that for the measure treewidth, which captures structural density of a program, the evaluation of the well-known class of normal programs is expected to be slightly harder than deciding satisfiability (SAT). However, it is unclear how to utilize this structural power of ASP. This paper deals with a novel reduction from SAT to normal ASP that goes beyond well-known encodings: We explicitly utilize the structural power of ASP, whereby we sublinearly decrease the treewidth, which probably cannot be significantly improved. Then, compared to existing results, this characterizes hardness in a fine-grained way by establishing the required functional dependency of the dependency graph’s cycle length (SCC size) on the treewidth. Thereby, we do not only improve this dependency for a family of almost-tight programs, we also utilize the reduction for lower bounds that close the gap to existing upper bounds.

Premium content

BridgeTower: Building Bridges Between Encoders in Vision-Language Representation Learning

Downloads

Next from AAAI 2023

Fair Short Paths in Vertex-Colored Graphs

Similar lecture

NLIP: Noise-robust Language-Image Pre-training