Search

We propose a video editing framework, NaRCan, which integrates a hybrid deformation field and diffusion prior to generate high-quality natural canonical images to represent the input video. Our approach utilizes homography to model global motion and employs multi-layer perceptrons (MLPs) to capture local residual deformations, enhancing the model's ability to handle complex video dynamics. By introducing a diffusion prior from the early stages of training, our model ensures that the generated images retain a high-quality natural appearance, making the produced canonical images suitable for various downstream tasks in video editing, a capability not achieved by current canonical-based methods. Furthermore, we incorporate low-rank adaptation (LoRA) fine-tuning and introduce a noise and diffusion prior update scheduling technique that accelerates the training process by 14 times. Extensive experimental results show that our method outperforms existing approaches in various video editing tasks and produces coherent and high-quality edited video sequences. See our project page for video results at https://koi953215.github.io/NaRCan_page/., Comment: Project page: https://koi953215.github.io/NaRCan_page/

This work investigates the oracle separation between the physically motivated complexity class of noisy quantum circuits, inspired by definitions such as those presented by Chen, Cotler, Huang, and Li (2022). We establish that with a constant error rate, separation can be achieved in terms of NP. When the error rate is $\Omega(\log n/n)$, we can extend this result to the separation of PH. Notably, our oracles, in all separations, do not necessitate error correction schemes or fault tolerance, as all quantum circuits are of constant depth. This indicates that even quantum computers with minor errors, without error correction, may surpass classical complexity classes under various scenarios and assumptions. We also explore various common noise settings and present new classical hardness results, generalizing those found in studies by Raz and Tal (2022) and Bassirian, Bouland, Fefferman, Gunn, and Tal (2021), which are of independent interest.

Quantum computing has emerged as a transformative tool for future data management. Classical problems in database domains, including query optimization, data integration, and transaction management, have recently been addressed using quantum computing techniques. This tutorial aims to establish the theoretical foundation essential for enhancing methodologies and practical implementations in this line of research. Moreover, this tutorial takes a forward-looking approach by delving into recent strides in quantum internet technologies and the nonlocality theory. We aim to shed light on the uncharted territory of future data systems tailored for the quantum internet., Comment: Accepted at the 40th IEEE International Conference on Data Engineering (ICDE 2024)

The report demonstrates the benefits (in terms of improved claims loss modeling) of harnessing the value of Federated Learning (FL) to learn a single model across multiple insurance industry datasets without requiring the datasets themselves to be shared from one company to another. The application of FL addresses two of the most pressing concerns: limited data volume and data variety, which are caused by privacy concerns, the rarity of claim events, the lack of informative rating factors, etc.. During each round of FL, collaborators compute improvements on the model using their local private data, and these insights are combined to update a global model. Such aggregation of insights allows for an increase to the effectiveness in forecasting claims losses compared to models individually trained at each collaborator. Critically, this approach enables machine learning collaboration without the need for raw data to leave the compute infrastructure of each respective data owner. Additionally, the open-source framework, OpenFL, that is used in our experiments is designed so that it can be run using confidential computing as well as with additional algorithmic protections against leakage of information via the shared model updates. In such a way, FL is implemented as a privacy-enhancing collaborative learning technique that addresses the challenges posed by the sensitivity and privacy of data in traditional machine learning solutions. This paper's application of FL can also be expanded to other areas including fraud detection, catastrophe modeling, etc., that have a similar need to incorporate data privacy into machine learning collaborations. Our framework and empirical results provide a foundation for future collaborations among insurers, regulators, academic researchers, and InsurTech experts.

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.

Lack of rigorous reproducibility and validation are major hurdles for scientific development across many fields. Materials science in particular encompasses a variety of experimental and theoretical approaches that require careful benchmarking. Leaderboard efforts have been developed previously to mitigate these issues. However, a comprehensive comparison and benchmarking on an integrated platform with multiple data modalities with both perfect and defect materials data is still lacking. This work introduces JARVIS-Leaderboard, an open-source and community-driven platform that facilitates benchmarking and enhances reproducibility. The platform allows users to set up benchmarks with custom tasks and enables contributions in the form of dataset, code, and meta-data submissions. We cover the following materials design categories: Artificial Intelligence (AI), Electronic Structure (ES), Force-fields (FF), Quantum Computation (QC) and Experiments (EXP). For AI, we cover several types of input data, including atomic structures, atomistic images, spectra, and text. For ES, we consider multiple ES approaches, software packages, pseudopotentials, materials, and properties, comparing results to experiment. For FF, we compare multiple approaches for material property predictions. For QC, we benchmark Hamiltonian simulations using various quantum algorithms and circuits. Finally, for experiments, we use the inter-laboratory approach to establish benchmarks. There are 1281 contributions to 274 benchmarks using 152 methods with more than 8 million data-points, and the leaderboard is continuously expanding. The JARVIS-Leaderboard is available at the website: https://pages.nist.gov/jarvis_leaderboard

Existing dense or paragraph video captioning approaches rely on holistic representations of videos, possibly coupled with learned object/action representations, to condition hierarchical language decoders. However, they fundamentally lack the commonsense knowledge of the world required to reason about progression of events, causality, and even the function of certain objects within a scene. To address this limitation we propose a novel video captioning Transformer-based model, that takes into account both implicit (visuo-lingual and purely linguistic) and explicit (knowledge-base) commonsense knowledge. We show that these forms of knowledge, in isolation and in combination, enhance the quality of produced captions. Further, inspired by imitation learning, we propose a new task of instruction generation, where the goal is to produce a set of linguistic instructions from a video demonstration of its performance. We formalize the task using the ALFRED dataset [54] generated using an AI2-THOR environment. While instruction generation is conceptually similar to paragraph captioning, it differs in the fact that it exhibits stronger object persistence, as well as spatially-aware and causal sentence structure. We show that our commonsense knowledge enhanced approach produces significant improvements on this task (up to 57% in METEOR and 8.5% in CIDEr), as well as the state-of-the-art result on more traditional video captioning in the ActivityNet Captions dataset [29]., Comment: The paper is under consideration at Computer Vision and Image Understanding Journal

We propose an application for near-term quantum devices: namely, generating cryptographically certified random bits, to use (for example) in proof-of-stake cryptocurrencies. Our protocol repurposes the existing "quantum supremacy" experiments, based on random circuit sampling, that Google and USTC have successfully carried out starting in 2019. We show that, whenever the outputs of these experiments pass the now-standard Linear Cross-Entropy Benchmark (LXEB), under plausible hardness assumptions they necessarily contain $\Omega(n)$ min-entropy, where $n$ is the number of qubits. To achieve a net gain in randomness, we use a small random seed to produce pseudorandom challenge circuits. In response to the challenge circuits, the quantum computer generates output strings that, after verification, can then be fed into a randomness extractor to produce certified nearly-uniform bits -- thereby "bootstrapping" from pseudorandomness to genuine randomness. We prove our protocol sound in two senses: (i) under a hardness assumption called Long List Quantum Supremacy Verification, which we justify in the random oracle model, and (ii) unconditionally in the random oracle model against an eavesdropper who could share arbitrary entanglement with the device. (Note that our protocol's output is unpredictable even to a computationally unbounded adversary who can see the random oracle.) Currently, the central drawback of our protocol is the exponential cost of verification, which in practice will limit its implementation to at most $n\sim 60$ qubits, a regime where attacks are expensive but not impossible. Modulo that drawback, our protocol appears to be the only practical application of quantum computing that both requires a QC and is physically realizable today., Comment: 84 pages

This paper explains the generalization power of a deep neural network (DNN) from the perspective of interactions. Although there is no universally accepted definition of the concepts encoded by a DNN, the sparsity of interactions in a DNN has been proved, i.e., the output score of a DNN can be well explained by a small number of interactions between input variables. In this way, to some extent, we can consider such interactions as interactive concepts encoded by the DNN. Therefore, in this paper, we derive an analytic explanation of inconsistency of concepts of different complexities. This may shed new lights on using the generalization power of concepts to explain the generalization power of the entire DNN. Besides, we discover that the DNN with stronger generalization power usually learns simple concepts more quickly and encodes fewer complex concepts. We also discover the detouring dynamics of learning complex concepts, which explains both the high learning difficulty and the low generalization power of complex concepts. The code will be released when the paper is accepted.

In this work, we study the computational complexity of quantum determinants, a $q$-deformation of matrix permanents: Given a complex number $q$ on the unit circle in the complex plane and an $n\times n$ matrix $X$, the $q$-permanent of $X$ is defined as $$\mathrm{Per}_q(X) = \sum_{\sigma\in S_n} q^{\ell(\sigma)}X_{1,\sigma(1)}\ldots X_{n,\sigma(n)},$$ where $\ell(\sigma)$ is the inversion number of permutation $\sigma$ in the symmetric group $S_n$ on $n$ elements. The function family generalizes determinant and permanent, which correspond to the cases $q=-1$ and $q=1$ respectively. For worst-case hardness, by Liouville's approximation theorem and facts from algebraic number theory, we show that for primitive $m$-th root of unity $q$ for odd prime power $m=p^k$, exactly computing $q$-permanent is $\mathsf{Mod}_p\mathsf{P}$-hard. This implies that an efficient algorithm for computing $q$-permanent results in a collapse of the polynomial hierarchy. Next, we show that computing $q$-permanent can be achieved using an oracle that approximates to within a polynomial multiplicative error and a membership oracle for a finite set of algebraic integers. From this, an efficient approximation algorithm would also imply a collapse of the polynomial hierarchy. By random self-reducibility, computing $q$-permanent remains to be hard for a wide range of distributions satisfying a property called the strong autocorrelation property. Specifically, this is proved via a reduction from $1$-permanent to $q$-permanent for $O(1/n^2)$ points $z$ on the unit circle. Since the family of permanent functions shares common algebraic structure, various techniques developed for the hardness of permanent can be generalized to $q$-permanents.

Deep learning models have been deployed in numerous real-world applications such as autonomous driving and surveillance. However, these models are vulnerable in adversarial environments. Backdoor attack is emerging as a severe security threat which injects a backdoor trigger into a small portion of training data such that the trained model behaves normally on benign inputs but gives incorrect predictions when the specific trigger appears. While most research in backdoor attacks focuses on image classification, backdoor attacks on object detection have not been explored but are of equal importance. Object detection has been adopted as an important module in various security-sensitive applications such as autonomous driving. Therefore, backdoor attacks on object detection could pose severe threats to human lives and properties. We propose four kinds of backdoor attacks for object detection task: 1) Object Generation Attack: a trigger can falsely generate an object of the target class; 2) Regional Misclassification Attack: a trigger can change the prediction of a surrounding object to the target class; 3) Global Misclassification Attack: a single trigger can change the predictions of all objects in an image to the target class; and 4) Object Disappearance Attack: a trigger can make the detector fail to detect the object of the target class. We develop appropriate metrics to evaluate the four backdoor attacks on object detection. We perform experiments using two typical object detection models -- Faster-RCNN and YOLOv3 on different datasets. More crucially, we demonstrate that even fine-tuning on another benign dataset cannot remove the backdoor hidden in the object detection model. To defend against these backdoor attacks, we propose Detector Cleanse, an entropy-based run-time detection framework to identify poisoned testing samples for any deployed object detector.

An engaging and provocative question can open up a great conversation. In this work, we explore a novel scenario: a conversation agent views a set of the user's photos (for example, from social media platforms) and asks an engaging question to initiate a conversation with the user. The existing vision-to-question models mostly generate tedious and obvious questions, which might not be ideals conversation starters. This paper introduces a two-phase framework that first generates a visual story for the photo set and then uses the story to produce an interesting question. The human evaluation shows that our framework generates more response-provoking questions for starting conversations than other vision-to-question baselines., Comment: Accepted as a full talk paper on AAAI-DEEPDIAL'21

We present two protocols for classical verification of quantum depth. Our protocols allow a purely classical verifier to distinguish devices with different quantum circuit depths even in the presence of classical computation. We show that a device with quantum circuit depth at most d will be rejected by the verifier even if the prover applies additional polynomial-time classical computation to cheat. On the other hand, the verifier accepts a device which has quantum circuit depth d' for some d'>d. In our first protocol, we introduce an additional untrusted quantum machine which shares entanglements with the target machine. Applying a robust self-test, our first protocol certifies the depth of the target machine with information theoretic security and nearly optimal separation. The protocol relies on the oracle separation problem for quantum depth by Chia, Chung and Lai [STOC 2020] and a transformation from an oracle separation problem to a two-player non-local game. Our second protocol certifies the quantum depth of a single device based on quantum hardness of learning with errors. The protocol relies on the noisy trapdoor claw-free function family and the idea of pointer chasing to force the prover to keep quantum coherence until all preceding message exchanges are completed. To our knowledge, we give the first constructions for distinguishing hybrid quantum-classical computers with different circuit depths in unrelativized models., Comment: 39 pages

This paper aims to theoretically analyze the complexity of feature transformations encoded in piecewise linear DNNs with ReLU layers. We propose metrics to measure three types of complexities of transformations based on the information theory. We further discover and prove the strong correlation between the complexity and the disentanglement of transformations. Based on the proposed metrics, we analyze two typical phenomena of the change of the transformation complexity during the training process, and explore the ceiling of a DNN's complexity. The proposed metrics can also be used as a loss to learn a DNN with the minimum complexity, which also controls the over-fitting level of the DNN and influences adversarial robustness, adversarial transferability, and knowledge consistency. Comprehensive comparative studies have provided new perspectives to understand the DNN.

Although machine learning (ML) has shown promise in numerous domains, there are concerns about generalizability to out-of-sample data. This is currently addressed by centrally sharing ample, and importantly diverse, data from multiple sites. However, such centralization is challenging to scale (or even not feasible) due to various limitations. Federated ML (FL) provides an alternative to train accurate and generalizable ML models, by only sharing numerical model updates. Here we present findings from the largest FL study to-date, involving data from 71 healthcare institutions across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, utilizing the largest dataset of such patients ever used in the literature (25,256 MRI scans from 6,314 patients). We demonstrate a 33% improvement over a publicly trained model to delineate the surgically targetable tumor, and 23% improvement over the tumor's entire extent. We anticipate our study to: 1) enable more studies in healthcare informed by large and diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further quantitative analyses for glioblastoma via performance optimization of our consensus model for eventual public release, and 3) demonstrate the effectiveness of FL at such scale and task complexity as a paradigm shift for multi-site collaborations, alleviating the need for data sharing., Comment: federated learning, deep learning, convolutional neural network, segmentation, brain tumor, glioma, glioblastoma, FeTS, BraTS