Cheng Zhang

Cheng Zhang, Xuanbai Chen, Siqi Chai et al. · cmu

Text-to-image generative models often reflect the biases of the training data, leading to unequal representations of underrepresented groups. This study investigates inclusive text-to-image generative models that generate images based on human-written prompts and ensure the resulting images are uniformly distributed across attributes of interest. Unfortunately, directly expressing the desired attributes in the prompt often leads to sub-optimal results due to linguistic ambiguity or model misrepresentation. Hence, this paper proposes a drastically different approach that adheres to the maxim that "a picture is worth a thousand words". We show that, for some attributes, images can represent concepts more expressively than text. For instance, categories of skin tones are typically hard to specify by text but can be easily represented by example images. Building upon these insights, we propose a novel approach, ITI-GEN, that leverages readily available reference images for Inclusive Text-to-Image GENeration. The key idea is learning a set of prompt embeddings to generate images that can effectively represent all desired attribute categories. More importantly, ITI-GEN requires no model fine-tuning, making it computationally efficient to augment existing text-to-image models. Extensive experiments demonstrate that ITI-GEN largely improves over state-of-the-art models to generate inclusive images from a prompt. Project page: https://czhang0528.github.io/iti-gen.

Cheng Zhang, Jian Shi, Xuan Deng et al.

In computer-aided design (CAD) community, the point cloud data is pervasively applied in reverse engineering, where the point cloud analysis plays an important role. While a large number of supervised learning methods have been proposed to handle the unordered point clouds and demonstrated their remarkable success, their performance and applicability are limited to the costly data annotation. In this work, we propose a novel self-supervised pretraining model for point cloud learning without human annotations, which relies solely on upsampling operation to perform feature learning of point cloud in an effective manner. The key premise of our approach is that upsampling operation encourages the network to capture both high-level semantic information and low-level geometric information of the point cloud, thus the downstream tasks such as classification and segmentation will benefit from the pre-trained model. Specifically, our method first conducts the random subsampling from the input point cloud at a low proportion e.g., 12.5%. Then, we feed them into an encoder-decoder architecture, where an encoder is devised to operate only on the subsampled points, along with a upsampling decoder is adopted to reconstruct the original point cloud based on the learned features. Finally, we design a novel joint loss function which enforces the upsampled points to be similar with the original point cloud and uniformly distributed on the underlying shape surface. By adopting the pre-trained encoder weights as initialisation of models for downstream tasks, we find that our UAE outperforms previous state-of-the-art methods in shape classification, part segmentation and point cloud upsampling tasks. Code will be made publicly available upon acceptance.

Zixi Zhang, Cheng Zhang, Xitong Gao et al.

Low-rank Adaption (LoRA) has been the de-facto parameter-efficient fine-tuning technique for large language models. We present HeteroLoRA, a light-weight search algorithm that leverages zero-cost proxies to allocate the limited LoRA trainable parameters across the model for better fine-tuned performance. In addition to the allocation for the standard LoRA-adapted models, we also demonstrate the efficacy of HeteroLoRA by performing the allocation in a more challenging search space that includes LoRA modules and LoRA-adapted shortcut connections. Experiments show that HeteroLoRA enables improvements in model performance given the same parameter budge. For example, on MRPC, we see an improvement of 1.6% in accuracy with similar training parameter budget. We will open-source our algorithm once the paper is accepted.

Jihyung Kil, Cheng Zhang, Dong Xuan et al.

Visual question answering (VQA) is challenging not only because the model has to handle multi-modal information, but also because it is just so hard to collect sufficient training examples -- there are too many questions one can ask about an image. As a result, a VQA model trained solely on human-annotated examples could easily over-fit specific question styles or image contents that are being asked, leaving the model largely ignorant about the sheer diversity of questions. Existing methods address this issue primarily by introducing an auxiliary task such as visual grounding, cycle consistency, or debiasing. In this paper, we take a drastically different approach. We found that many of the "unknowns" to the learned VQA model are indeed "known" in the dataset implicitly. For instance, questions asking about the same object in different images are likely paraphrases; the number of detected or annotated objects in an image already provides the answer to the "how many" question, even if the question has not been annotated for that image. Building upon these insights, we present a simple data augmentation pipeline SimpleAug to turn this "known" knowledge into training examples for VQA. We show that these augmented examples can notably improve the learned VQA models' performance, not only on the VQA-CP dataset with language prior shifts but also on the VQA v2 dataset without such shifts. Our method further opens up the door to leverage weakly-labeled or unlabeled images in a principled way to enhance VQA models. Our code and data are publicly available at https://github.com/heendung/simpleAUG.

Shih-Han Chou, Matthew Kowal, Yasmin Niknam et al.

While progress has been made in the domain of video-language understanding, current state-of-the-art algorithms are still limited in their ability to understand videos at high levels of abstraction, such as news-oriented videos. Alternatively, humans easily amalgamate information from video and language to infer information beyond what is visually observable in the pixels. An example of this is watching a news story, where the context of the event can play as big of a role in understanding the story as the event itself. Towards a solution for designing this ability in algorithms, we present a large-scale analysis on an in-house dataset collected by the Reuters News Agency, called Reuters Video-Language News (ReutersViLNews) dataset which focuses on high-level video-language understanding with an emphasis on long-form news. The ReutersViLNews Dataset consists of long-form news videos collected and labeled by news industry professionals over several years and contains prominent news reporting from around the world. Each video involves a single story and contains action shots of the actual event, interviews with people associated with the event, footage from nearby areas, and more. ReutersViLNews dataset contains videos from seven subject categories: disaster, finance, entertainment, health, politics, sports, and miscellaneous with annotations from high-level to low-level, title caption, visual video description, high-level story description, keywords, and location. We first present an analysis of the dataset statistics of ReutersViLNews compared to previous datasets. Then we benchmark state-of-the-art approaches for four different video-language tasks. The results suggest that news-oriented videos are a substantial challenge for current video-language understanding algorithms and we conclude by providing future directions in designing approaches to solve the ReutersViLNews dataset.

Cheng Zhang, Haocheng Wan, Xinyi Shen et al.

The recently developed pure Transformer architectures have attained promising accuracy on point cloud learning benchmarks compared to convolutional neural networks. However, existing point cloud Transformers are computationally expensive since they waste a significant amount of time on structuring the irregular data. To solve this shortcoming, we present Sparse Window Attention (SWA) module to gather coarse-grained local features from non-empty voxels, which not only bypasses the expensive irregular data structuring and invalid empty voxel computation, but also obtains linear computational complexity with respect to voxel resolution. Meanwhile, to gather fine-grained features about the global shape, we introduce relative attention (RA) module, a more robust self-attention variant for rigid transformations of objects. Equipped with the SWA and RA, we construct our neural architecture called PVT that integrates both modules into a joint framework for point cloud learning. Compared with previous Transformer-based and attention-based models, our method attains top accuracy of 94.0% on classification benchmark and 10x inference speedup on average. Extensive experiments also valid the effectiveness of PVT on part and semantic segmentation benchmarks (86.6% and 69.2% mIoU, respectively).

Cheng Zhang, Jinwoo Kim, JungHo Jeon et al.

The purpose of this paper is to examine the opportunities and barriers of Integrated Human-Machine Intelligence (IHMI) in civil engineering. Integrating artificial intelligence's high efficiency and repeatability with humans' adaptability in various contexts can advance timely and reliable decision-making during civil engineering projects and emergencies. Successful cases in other domains, such as biomedical science, healthcare, and transportation, showed the potential of IHMI in data-driven, knowledge-based decision-making in numerous civil engineering applications. However, whether the industry and academia are ready to embrace the era of IHMI and maximize its benefit to the industry is still questionable due to several knowledge gaps. This paper thus calls for future studies in exploring the value, method, and challenges of applying IHMI in civil engineering. Our systematic review of the literature and motivating cases has identified four knowledge gaps in achieving effective IHMI in civil engineering. First, it is unknown what types of tasks in the civil engineering domain can be assisted by AI and to what extent. Second, the interface between human and AI in civil engineering-related tasks need more precise and formal definition. Third, the barriers that impede collecting detailed behavioral data from humans and contextual environments deserve systematic classification and prototyping. Lastly, it is unknown what expected and unexpected impacts will IHMI have on the AEC industry and entrepreneurship. Analyzing these knowledge gaps led to a list of identified research questions. This paper will lay the foundation for identifying relevant studies to form a research roadmap to address the four knowledge gaps identified.

Angus Lamb, Evgeny Saveliev, Yingzhen Li et al.

While deep learning has obtained state-of-the-art results in many applications, the adaptation of neural network architectures to incorporate new output features remains a challenge, as neural networks are commonly trained to produce a fixed output dimension. This issue is particularly severe in online learning settings, where new output features, such as items in a recommender system, are added continually with few or no associated observations. As such, methods for adapting neural networks to novel features which are both time and data-efficient are desired. To address this, we propose the Contextual HyperNetwork (CHN), an auxiliary model which generates parameters for extending the base model to a new feature, by utilizing both existing data as well as any observations and/or metadata associated with the new feature. At prediction time, the CHN requires only a single forward pass through a neural network, yielding a significant speed-up when compared to re-training and fine-tuning approaches. To assess the performance of CHNs, we use a CHN to augment a partial variational autoencoder (P-VAE), a deep generative model which can impute the values of missing features in sparsely-observed data. We show that this system obtains improved few-shot learning performance for novel features over existing imputation and meta-learning baselines across recommender systems, e-learning, and healthcare tasks.

Zichao Wang, Angus Lamb, Evgeny Saveliev et al.

This competition concerns educational diagnostic questions, which are pedagogically effective, multiple-choice questions (MCQs) whose distractors embody misconceptions. With a large and ever-increasing number of such questions, it becomes overwhelming for teachers to know which questions are the best ones to use for their students. We thus seek to answer the following question: how can we use data on hundreds of millions of answers to MCQs to drive automatic personalized learning in large-scale learning scenarios where manual personalization is infeasible? Success in using MCQ data at scale helps build more intelligent, personalized learning platforms that ultimately improve the quality of education en masse. To this end, we introduce a new, large-scale, real-world dataset and formulate 4 data mining tasks on MCQs that mimic real learning scenarios and target various aspects of the above question in a competition setting at NeurIPS 2020. We report on our NeurIPS competition in which nearly 400 teams submitted approximately 4000 submissions, with encouragingly diverse and effective approaches to each of our tasks.

Yongming Huang, Shengheng Liu, Cheng Zhang et al.

Future beyond fifth-generation (B5G) and sixth-generation (6G) mobile communications will shift from facilitating interpersonal communications to supporting Internet of Everything (IoE), where intelligent communications with full integration of big data and artificial intelligence (AI) will play an important role in improving network efficiency and providing high-quality service. As a rapid evolving paradigm, the AI-empowered mobile communications demand large amounts of data acquired from real network environment for systematic test and verification. Hence, we build the world's first true-data testbed for 5G/B5G intelligent network (TTIN), which comprises 5G/B5G on-site experimental networks, data acquisition & data warehouse, and AI engine & network optimization. In the TTIN, true network data acquisition, storage, standardization, and analysis are available, which enable system-level online verification of B5G/6G-orientated key technologies and support data-driven network optimization through the closed-loop control mechanism. This paper elaborates on the system architecture and module design of TTIN. Detailed technical specifications and some of the established use cases are also showcased.

Cheng Zhang, Wei-Lun Chao, Dong Xuan

Visual question answering (Visual QA) has attracted significant attention these years. While a variety of algorithms have been proposed, most of them are built upon different combinations of image and language features as well as multi-modal attention and fusion. In this paper, we investigate an alternative approach inspired by conventional QA systems that operate on knowledge graphs. Specifically, we investigate the use of scene graphs derived from images for Visual QA: an image is abstractly represented by a graph with nodes corresponding to object entities and edges to object relationships. We adapt the recently proposed graph network (GN) to encode the scene graph and perform structured reasoning according to the input question. Our empirical studies demonstrate that scene graphs can already capture essential information of images and graph networks have the potential to outperform state-of-the-art Visual QA algorithms but with a much cleaner architecture. By analyzing the features generated by GNs we can further interpret the reasoning process, suggesting a promising direction towards explainable Visual QA.