arXiv subjects

Ting Chen

Publications and source records attributed to Ting Chen.

At least 109 records · Page 6Linked to original sources

arXiv2022

Robust and Efficient Medical Imaging with Self-Supervision

Recent progress in Medical Artificial Intelligence (AI) has delivered systems that can reach clinical expert level performance. However, such systems tend to demonstrate sub-optimal "out-of-distribution" performance when evaluated in clinical settings different from the training environment. A common mitigation strategy is to develop separate systems for each clinical setting using site-specific data [1]. However, this quickly becomes impractical as medical data is time-consuming to acquire and expensive to annotate [2]. Thus, the problem of "data-efficient generalization" presents an ongoing difficulty for Medical AI development. Although progress in representation learning shows promise, their benefits have not been rigorously studied, specifically for out-of-distribution settings. To meet these challenges, we present REMEDIS, a unified representation learning strategy to improve robustness and data-efficiency of medical imaging AI. REMEDIS uses a generic combination of large-scale supervised transfer learning with self-supervised learning and requires little task-specific customization. We study a diverse range of medical imaging tasks and simulate three realistic application scenarios using retrospective data. REMEDIS exhibits significantly improved in-distribution performance with up to 11.5% relative improvement in diagnostic accuracy over a strong supervised baseline. More importantly, our strategy leads to strong data-efficient generalization of medical imaging AI, matching strong supervised baselines using between 1% to 33% of retraining data across tasks. These results suggest that REMEDIS can significantly accelerate the life-cycle of medical imaging AI development thereby presenting an important step forward for medical imaging AI to deliver broad impact.

cs.CV↗

arXiv2022

Scalable Online Disease Diagnosis via Multi-Model-Fused Actor-Critic Reinforcement Learning

For those seeking healthcare advice online, AI based dialogue agents capable of interacting with patients to perform automatic disease diagnosis are a viable option. This application necessitates efficient inquiry of relevant disease symptoms in order to make accurate diagnosis recommendations. This can be formulated as a problem of sequential feature (symptom) selection and classification for which reinforcement learning (RL) approaches have been proposed as a natural solution. They perform well when the feature space is small, that is, the number of symptoms and diagnosable disease categories is limited, but they frequently fail in assignments with a large number of features. To address this challenge, we propose a Multi-Model-Fused Actor-Critic (MMF-AC) RL framework that consists of a generative actor network and a diagnostic critic network. The actor incorporates a Variational AutoEncoder (VAE) to model the uncertainty induced by partial observations of features, thereby facilitating in making appropriate inquiries. In the critic network, a supervised diagnosis model for disease predictions is involved to precisely estimate the state-value function. Furthermore, inspired by the medical concept of differential diagnosis, we combine the generative and diagnosis models to create a novel reward shaping mechanism to address the sparse reward problem in large search spaces. We conduct extensive experiments on both synthetic and real-world datasets for empirical evaluations. The results demonstrate that our approach outperforms state-of-the-art methods in terms of diagnostic accuracy and interaction efficiency while also being more effectively scalable to large search spaces. Besides, our method is adaptable to both categorical and continuous features, making it ideal for online applications.

cs.LG↗

arXiv2022

Decoder Denoising Pretraining for Semantic Segmentation

Semantic segmentation labels are expensive and time consuming to acquire. Hence, pretraining is commonly used to improve the label-efficiency of segmentation models. Typically, the encoder of a segmentation model is pretrained as a classifier and the decoder is randomly initialized. Here, we argue that random initialization of the decoder can be suboptimal, especially when few labeled examples are available. We propose a decoder pretraining approach based on denoising, which can be combined with supervised pretraining of the encoder. We find that decoder denoising pretraining on the ImageNet dataset strongly outperforms encoder-only supervised pretraining. Despite its simplicity, decoder denoising pretraining achieves state-of-the-art results on label-efficient semantic segmentation and offers considerable gains on the Cityscapes, Pascal Context, and ADE20K datasets.

cs.CV↗

arXiv2022

Investigation of the Effect of Quantum Measurement on Parity-Time Symmetry

Symmetry, including the parity-time ($\mathcal{PT}$)-symmetry, is a striking topic, widely discussed and employed in many fields. It is well-known that quantum measurement can destroy or disturb quantum systems. However, can and how does quantum measurement destroy the symmetry of the measured system? To answer the pertinent question, we establish the correlation between the quantum measurement and Floquet $\mathcal{PT}$-symmetry and investigate for the first time how the measurement frequency and measurement strength affect the $\mathcal{PT}$-symmetry of the measured system using the $^{40}\mathrm{Ca}^{+}$ ion. It is already shown that the measurement at high frequencies would break the $\mathcal{PT}$ symmetry. Notably, even for an inadequately fast measurement frequency, if the measurement strength is sufficiently strong, the $\mathcal{PT}$ symmetry breaking can occur. The current work can enhance our knowledge of quantum measurement and symmetry and may inspire further research on the effect of quantum measurement on symmetry.

quant-ph↗

arXiv2022

Pix2seq: A Language Modeling Framework for Object Detection

We present Pix2Seq, a simple and generic framework for object detection. Unlike existing approaches that explicitly integrate prior knowledge about the task, we cast object detection as a language modeling task conditioned on the observed pixel inputs. Object descriptions (e.g., bounding boxes and class labels) are expressed as sequences of discrete tokens, and we train a neural network to perceive the image and generate the desired sequence. Our approach is based mainly on the intuition that if a neural network knows about where and what the objects are, we just need to teach it how to read them out. Beyond the use of task-specific data augmentations, our approach makes minimal assumptions about the task, yet it achieves competitive results on the challenging COCO dataset, compared to highly specialized and well optimized detection algorithms.

cs.CV↗

arXiv2022

SNPSFuzzer: A Fast Greybox Fuzzer for Stateful Network Protocols using Snapshots

Greybox fuzzing has been widely used in stateless programs and has achieved great success. However, most state-of-the-art greybox fuzzers generally have the problems of slow speed and shallow state depth coverage in the process of fuzzing stateful network protocol programs which are able to remember and store details of the interactions. The existing greybox fuzzers for network protocol programs send a series of well-defined prefix sequences of input messages first and then send mutated messages to test the target state of a stateful network protocol. The process mentioned above causes a high time cost. In this paper, we propose SNPSFuzzer, a fast greybox fuzzer for stateful network protocol using snapshots. SNPSFuzzer dumps the context information when the network protocol program is under a specific state and restores it when the state needs to be fuzzed. Furthermore, we design a message chain analysis algorithm to explore more and deeper network protocol states. Our evaluation shows that, compared with the state-of-the-art network protocol greybox fuzzer AFLNET, SNPSFuzzer increases the speed of network protocol fuzzing by 112.0%-168.9% and improves path coverage by 21.4%-27.5% within 24 hours. Moreover, SNPSFuzzer exposes a previously unreported vulnerability in program Tinydtls.

cs.CR↗

arXiv2022

BSODA: A Bipartite Scalable Framework for Online Disease Diagnosis

A growing number of people are seeking healthcare advice online. Usually, they diagnose their medical conditions based on the symptoms they are experiencing, which is also known as self-diagnosis. From the machine learning perspective, online disease diagnosis is a sequential feature (symptom) selection and classification problem. Reinforcement learning (RL) methods are the standard approaches to this type of tasks. Generally, they perform well when the feature space is small, but frequently become inefficient in tasks with a large number of features, such as the self-diagnosis. To address the challenge, we propose a non-RL Bipartite Scalable framework for Online Disease diAgnosis, called BSODA. BSODA is composed of two cooperative branches that handle symptom-inquiry and disease-diagnosis, respectively. The inquiry branch determines which symptom to collect next by an information-theoretic reward. We employ a Product-of-Experts encoder to significantly improve the handling of partial observations of a large number of features. Besides, we propose several approximation methods to substantially reduce the computational cost of the reward to a level that is acceptable for online services. Additionally, we leverage the diagnosis model to estimate the reward more precisely. For the diagnosis branch, we use a knowledge-guided self-attention model to perform predictions. In particular, BSODA determines when to stop inquiry and output predictions using both the inquiry and diagnosis models. We demonstrate that BSODA outperforms the state-of-the-art methods on several public datasets. Moreover, we propose a novel evaluation method to test the transferability of symptom checking methods from synthetic to real-world tasks. Compared to existing RL baselines, BSODA is more effectively scalable to large search spaces.

cs.LG↗

arXiv2021

Nested Hierarchical Transformer: Towards Accurate, Data-Efficient and Interpretable Visual Understanding

Hierarchical structures are popular in recent vision transformers, however, they require sophisticated designs and massive datasets to work well. In this paper, we explore the idea of nesting basic local transformers on non-overlapping image blocks and aggregating them in a hierarchical way. We find that the block aggregation function plays a critical role in enabling cross-block non-local information communication. This observation leads us to design a simplified architecture that requires minor code changes upon the original vision transformer. The benefits of the proposed judiciously-selected design are threefold: (1) NesT converges faster and requires much less training data to achieve good generalization on both ImageNet and small datasets like CIFAR; (2) when extending our key ideas to image generation, NesT leads to a strong decoder that is 8$\times$ faster than previous transformer-based generators; and (3) we show that decoupling the feature learning and abstraction processes via this nested hierarchy in our design enables constructing a novel method (named GradCAT) for visually interpreting the learned model. Source code is available https://github.com/google-research/nested-transformer.

cs.CV↗

arXiv2021

Improved Transformer for High-Resolution GANs

Attention-based models, exemplified by the Transformer, can effectively model long range dependency, but suffer from the quadratic complexity of self-attention operation, making them difficult to be adopted for high-resolution image generation based on Generative Adversarial Networks (GANs). In this paper, we introduce two key ingredients to Transformer to address this challenge. First, in low-resolution stages of the generative process, standard global self-attention is replaced with the proposed multi-axis blocked self-attention which allows efficient mixing of local and global attention. Second, in high-resolution stages, we drop self-attention while only keeping multi-layer perceptrons reminiscent of the implicit neural function. To further improve the performance, we introduce an additional self-modulation component based on cross-attention. The resulting model, denoted as HiT, has a nearly linear computational complexity with respect to the image size and thus directly scales to synthesizing high definition images. We show in the experiments that the proposed HiT achieves state-of-the-art FID scores of 30.83 and 2.95 on unconditional ImageNet $128 \times 128$ and FFHQ $256 \times 256$, respectively, with a reasonable throughput. We believe the proposed HiT is an important milestone for generators in GANs which are completely free of convolutions. Our code is made publicly available at https://github.com/google-research/hit-gan

cs.CV↗

arXiv2021

Improving Contrastive Learning on Imbalanced Seed Data via Open-World Sampling

Contrastive learning approaches have achieved great success in learning visual representations with few labels of the target classes. That implies a tantalizing possibility of scaling them up beyond a curated "seed" benchmark, to incorporating more unlabeled images from the internet-scale external sources to enhance its performance. However, in practice, larger amount of unlabeled data will require more computing resources due to the bigger model size and longer training needed. Moreover, open-world unlabeled data usually follows an implicit long-tail class or attribute distribution, many of which also do not belong to the target classes. Blindly leveraging all unlabeled data hence can lead to the data imbalance as well as distraction issues. This motivates us to seek a principled approach to strategically select unlabeled data from an external source, in order to learn generalizable, balanced and diverse representations for relevant classes. In this work, we present an open-world unlabeled data sampling framework called Model-Aware K-center (MAK), which follows three simple principles: (1) tailness, which encourages sampling of examples from tail classes, by sorting the empirical contrastive loss expectation (ECLE) of samples over random data augmentations; (2) proximity, which rejects the out-of-distribution outliers that may distract training; and (3) diversity, which ensures diversity in the set of sampled examples. Empirically, using ImageNet-100-LT (without labels) as the seed dataset and two "noisy" external data sources, we demonstrate that MAK can consistently improve both the overall representation quality and the class balancedness of the learned features, as evaluated via linear classifier evaluation on full-shot and few-shot settings. The code is available at: https://github.com/VITA-Group/MAK

cs.CV↗

arXiv2021

Parallel-Electromagnetically-Induced-Transparency Near Ground-State Cooling of a Trapped-ion Crystal

We theoretically propose and experimentally demonstrate a parallel-electromagnetically-induced transparency (parallel-EIT) cooling technique for ion crystals in the Paul trap. It has less stringent requirements on the cooling resonance condition than the standard electromagnetically-induced transparency (EIT) cooling, thus allowing, in principle, to simultaneously cool the motional mode spectrum with an arbitrary range. A proof-of-principle validation for this cooling scheme is experimentally demonstrated with up to 4 trapped 40Ca+ ions. We observe simultaneous near-ground-state cooling for all motional modes with best average phonon number about 0.2. By tuning the trap frequency in a large range to imitate a broadband motional mode spectrum, we can still reach almost the same cooling limit for all the modes while standard EIT cooling shows limited cooling range. Our method has a simple experimental configuration, requiring only appropriate modulation of the probe beam of standard EIT cooling, and can be applied to various types of ions (e.g., 171Yb+, 40Ca+). This cooling scheme provides a powerful tool for the initialization of the trapped-ion quantum computers and simulators.

quant-ph↗

arXiv2021

TiKick: Towards Playing Multi-agent Football Full Games from Single-agent Demonstrations

Deep reinforcement learning (DRL) has achieved super-human performance on complex video games (e.g., StarCraft II and Dota II). However, current DRL systems still suffer from challenges of multi-agent coordination, sparse rewards, stochastic environments, etc. In seeking to address these challenges, we employ a football video game, e.g., Google Research Football (GRF), as our testbed and develop an end-to-end learning-based AI system (denoted as TiKick) to complete this challenging task. In this work, we first generated a large replay dataset from the self-playing of single-agent experts, which are obtained from league training. We then developed a distributed learning system and new offline algorithms to learn a powerful multi-agent AI from the fixed single-agent dataset. To the best of our knowledge, Tikick is the first learning-based AI system that can take over the multi-agent Google Research Football full game, while previous work could either control a single agent or experiment on toy academic scenarios. Extensive experiments further show that our pre-trained model can accelerate the training process of the modern multi-agent algorithm and our method achieves state-of-the-art performances on various academic scenarios.

cs.AI↗

arXiv2021

Nucleon isovector scalar charge from overlap fermions

We calculate the nucleon isovector scalar charge in lattice QCD using overlap fermions on five ensembles of gauge configurations generated by the RBC/UKQCD collaboration using the domain-wall quark action with $2+1$ dynamical flavors. The five ensembles cover five pion masses, $m_π\approx$ 139, 171, 302, 337 and 371 MeV, and four lattice spacings, $a \approx $ 0.06, 0.08, 0.11 and 0.14 fm. Three to six valence quark masses are computed on each ensemble to investigate the pion mass dependence. The extrapolation to the physical pion mass, continuum and infinite volume limits is obtained by a global fit of all data to a formula originated from partially quenched chiral perturbation theory. The excited-states contamination is carefully analyzed with 3--5 sink-source separations and multi-state fits. Our final result, in the $\overline{\text{MS}}$ scheme at 2 GeV, is $g_{S}^{u-d}= 0.94 (10)_{stat}(8)_{sys}$, where the first error is the statistical error and the second is the systematic error.

hep-lat↗

arXiv2021

Why Do Better Loss Functions Lead to Less Transferable Features?

Previous work has proposed many new loss functions and regularizers that improve test accuracy on image classification tasks. However, it is not clear whether these loss functions learn better representations for downstream tasks. This paper studies how the choice of training objective affects the transferability of the hidden representations of convolutional neural networks trained on ImageNet. We show that many objectives lead to statistically significant improvements in ImageNet accuracy over vanilla softmax cross-entropy, but the resulting fixed feature extractors transfer substantially worse to downstream tasks, and the choice of loss has little effect when networks are fully fine-tuned on the new tasks. Using centered kernel alignment to measure similarity between hidden representations of networks, we find that differences among loss functions are apparent only in the last few layers of the network. We delve deeper into representations of the penultimate layer, finding that different objectives and hyperparameter combinations lead to dramatically different levels of class separation. Representations with higher class separation obtain higher accuracy on the original task, but their features are less useful for downstream tasks. Our results suggest there exists a trade-off between learning invariant features for the original task and features relevant for transfer tasks.

cs.CV↗

arXiv2021

Intriguing Properties of Contrastive Losses

We study three intriguing properties of contrastive learning. First, we generalize the standard contrastive loss to a broader family of losses, and we find that various instantiations of the generalized loss perform similarly under the presence of a multi-layer non-linear projection head. Second, we study if instance-based contrastive learning (with a global image representation) can learn well on images with multiple objects present. We find that meaningful hierarchical local features can be learned despite the fact that these objectives operate on global instance-level features. Finally, we study the phenomenon of feature suppression among competing features shared across augmented views, such as "color distribution" vs "object class". We construct datasets with explicit and controllable competing features, and show that, for contrastive learning, a few bits of easy-to-learn shared features can suppress, and even fully prevent, the learning of other sets of competing features. In scenarios where there are multiple objects in an image, the dominant object would suppress the learning of smaller objects. Existing contrastive learning methods critically rely on data augmentation to favor certain sets of features over others, and could suffer from learning saturation for scenarios where existing augmentations cannot fully address the feature suppression. This poses open challenges to existing contrastive learning techniques.

cs.LG↗

arXiv2021

Ranking Cost: Building An Efficient and Scalable Circuit Routing Planner with Evolution-Based Optimization

Circuit routing has been a historically challenging problem in designing electronic systems such as very large-scale integration (VLSI) and printed circuit boards (PCBs). The main challenge is that connecting a large number of electronic components under specific design rules involves a very large search space. Early solutions are typically designed with hard-coded heuristics, which suffer from problems of non-optimal solutions and lack of flexibility for new design needs. Although a few learning-based methods have been proposed recently, they are typically cumbersome and hard to extend to large-scale applications. In this work, we propose a new algorithm for circuit routing, named Ranking Cost, which innovatively combines search-based methods (i.e., A* algorithm) and learning-based methods (i.e., Evolution Strategies) to form an efficient and trainable router. In our method, we introduce a new set of variables called cost maps, which can help the A* router to find out proper paths to achieve the global objective. We also train a ranking parameter, which can produce the ranking order and further improve the performance of our method. Our algorithm is trained in an end-to-end manner and does not use any artificial data or human demonstration. In the experiments, we compare with the sequential A* algorithm and a canonical reinforcement learning approach, and results show that our method outperforms these baselines with higher connectivity rates and better scalability.

cs.AI↗

arXiv2021

MURAL: Multimodal, Multitask Retrieval Across Languages

Both image-caption pairs and translation pairs provide the means to learn deep representations of and connections between languages. We use both types of pairs in MURAL (MUltimodal, MUltitask Representations Across Languages), a dual encoder that solves two tasks: 1) image-text matching and 2) translation pair matching. By incorporating billions of translation pairs, MURAL extends ALIGN (Jia et al. PMLR'21)--a state-of-the-art dual encoder learned from 1.8 billion noisy image-text pairs. When using the same encoders, MURAL's performance matches or exceeds ALIGN's cross-modal retrieval performance on well-resourced languages across several datasets. More importantly, it considerably improves performance on under-resourced languages, showing that text-text learning can overcome a paucity of image-caption examples for these languages. On the Wikipedia Image-Text dataset, for example, MURAL-base improves zero-shot mean recall by 8.1% on average for eight under-resourced languages and by 6.8% on average when fine-tuning. We additionally show that MURAL's text representations cluster not only with respect to genealogical connections but also based on areal linguistics, such as the Balkan Sprachbund.

cs.IR↗

arXiv2021

iBatch: Saving Ethereum Fees via Secure and Cost-Effective Batching of Smart-Contract Invocations

This paper presents iBatch, a middleware system running on top of an operational Ethereum network to enable secure batching of smart-contract invocations against an untrusted relay server off-chain. iBatch does so at a low overhead by validating the server's batched invocations in smart contracts without additional states. The iBatch mechanism supports a variety of policies, ranging from conservative to aggressive batching, and can be configured adaptively to the current workloads. iBatch automatically rewrites smart contracts to integrate with legacy applications and support large-scale deployment. For cost evaluation, we develop a platform with fast and cost-accurate transaction replaying, build real transaction benchmarks on popular Ethereum applications, and build a functional prototype of iBatch on Ethereum. The evaluation results show that iBatch saves 14.6%-59.1% Gas cost per invocation with a moderate 2-minute delay and 19.06%-31.52% Ether cost per invocation with a delay of 0.26-1.66 blocks.

cs.CR↗