Least Square Value Iteration is Robust Under Locally Bounded Misspecification Error

The success of reinforcement learning heavily relies on the function approximation of policy, value or models, where misspecification (a mismatch between the ground-truth and best function approximators) naturally occurs especially when the ground-truth is complex. As misspecification error does not vanish even with infinite number of samples, designing algorithms that are robust under misspecification is of paramount importance. Recently, it is shown that policy-based approaches can be robust even when the policy function approximation is under a large locally-bounded misspecification error, with which the function class may have $\Omega(1)$ approximation error in certain states and actions but is only small on average under a policy-induced state-distribution; whereas it is only known that value-based approach can effectively learn under globally-bounded misspecification error, i.e., the approximation errors to value functions have a uniform upper bound on all state-actions. Yet it remains an open question whether similar robustness can be achieved with value-based approaches. In this paper, we answer this question affirmatively by showing that the algorithm, Least-Square-Value-Iteration [Jin et al, 2020], with carefully designed exploration bonus can achieve robustness under local misspecification error bound. In particular, we show that algorithm achieves a regret bound of $\widetilde{O}\left(\sqrt{d^3KH^4} + dKH^2\zeta \right)$, where $d$ is the dimension of linear features, $H$ is the length of the episode, $K$ is the total number of episodes, and $\zeta$ is the local bound of the misspecification error. Moreover, we show that the algorithm can achieve the same regret bound without knowing $\zeta$ and can be used as robust policy evaluation oracle that can be applied to improve sample complexity in policy-based approaches.

Low-Switching Policy Gradient with Exploration via Online Sensitivity Sampling

Policy optimization methods are powerful algorithms in Reinforcement Learning (RL) for their flexibility to deal with policy parameterization and ability to handle model misspecification. However, these methods usually suffer from slow convergence rates and poor sample complexity. Hence it is important to design provably sample efficient algorithms for policy optimization. Yet, recent advances for this problems have only been successful in tabular and linear setting, whose benign structures cannot be generalized to non-linearly parameterized policies. In this paper, we address this problem by leveraging recent advances in value-based algorithms, including bounded eluder-dimension and online sensitivity sampling, to design a low-switching sample-efficient policy optimization algorithm, LPO, with general non-linear function approximation. We show that, our algorithm obtains an $\varepsilon$-optimal policy with only $\widetilde{O}(\frac{\text{poly}(d)}{\varepsilon^3})$ samples, where $\varepsilon$ is the suboptimality gap and $d$ is a complexity measure of the function class approximating the policy. This drastically improves previously best-known sample bound for policy optimization algorithms, $\widetilde{O}(\frac{\text{poly}(d)}{\varepsilon^8})$. Moreover, we empirically test our theory with deep neural nets to show the benefits of the theoretical inspiration.

Semi-supervised Cell Recognition under Point Supervision

Cell recognition is a fundamental task in digital histopathology image analysis. Point-based cell recognition (PCR) methods normally require a vast number of annotations, which is extremely costly, time-consuming and labor-intensive. Semi-supervised learning (SSL) can provide a shortcut to make full use of cell information in gigapixel whole slide images without exhaustive labeling. However, research into semi-supervised point-based cell recognition (SSPCR) remains largely overlooked. Previous SSPCR works are all built on density map-based PCR models, which suffer from unsatisfactory accuracy, slow inference speed and high sensitivity to hyper-parameters. To address these issues, end-to-end PCR models are proposed recently. In this paper, we develop a SSPCR framework suitable for the end-to-end PCR models for the first time. Overall, we use the current models to generate pseudo labels for unlabeled images, which are in turn utilized to supervise the models training. Besides, we introduce a co-teaching strategy to overcome the confirmation bias problem that generally exists in self-training. A distribution alignment technique is also incorporated to produce high-quality, unbiased pseudo labels for unlabeled data. Experimental results on four histopathology datasets concerning different types of staining styles show the effectiveness and versatility of the proposed framework. Code is available at \textcolor{magenta}{\url{https://github.com/windygooo/SSPCR}

PathAsst: Redefining Pathology through Generative Foundation AI Assistant for Pathology

As advances in large language models (LLMs) and multimodal techniques continue to mature, the development of general-purpose multimodal large language models (MLLMs) has surged, with significant applications in natural image interpretation. However, the field of pathology has largely remained untapped in this regard, despite the growing need for accurate, timely, and personalized diagnostics. To bridge the gap in pathology MLLMs, we present the PathAsst in this study, which is a generative foundation AI assistant to revolutionize diagnostic and predictive analytics in pathology. To develop PathAsst, we collect over 142K high-quality pathology image-text pairs from a variety of reliable sources, including PubMed, comprehensive pathology textbooks, reputable pathology websites, and private data annotated by pathologists. Leveraging the advanced capabilities of ChatGPT/GPT-4, we generate over 180K instruction-following samples. Furthermore, we devise additional instruction-following data, specifically tailored for the invocation of the pathology-specific models, allowing the PathAsst to effectively interact with these models based on the input image and user intent, consequently enhancing the model's diagnostic capabilities. Subsequently, our PathAsst is trained based on Vicuna-13B language model in coordination with the CLIP vision encoder. The results of PathAsst show the potential of harnessing the AI-powered generative foundation model to improve pathology diagnosis and treatment processes. We are committed to open-sourcing our meticulously curated dataset, as well as a comprehensive toolkit designed to aid researchers in the extensive collection and preprocessing of their own datasets. Resources can be obtained at https://github.com/superjamessyx/Generative-Foundation-AI-Assistant-for-Pathology.

MEDNC: Multi-ensemble deep neural network for COVID-19 diagnosis

Coronavirus disease 2019 (COVID-19) has spread all over the world for three years, but medical facilities in many areas still aren't adequate. There is a need for rapid COVID-19 diagnosis to identify high-risk patients and maximize the use of limited medical resources. Motivated by this fact, we proposed the deep learning framework MEDNC for automatic prediction and diagnosis of COVID-19 using computed tomography (CT) images. Our model was trained using two publicly available sets of COVID-19 data. And it was built with the inspiration of transfer learning. Results indicated that the MEDNC greatly enhanced the detection of COVID-19 infections, reaching an accuracy of 98.79% and 99.82% respectively. We tested MEDNC on a brain tumor and a blood cell dataset to show that our model applies to a wide range of problems. The outcomes demonstrated that our proposed models attained an accuracy of 99.39% and 99.28%, respectively. This COVID-19 recognition tool could help optimize healthcare resources and reduce clinicians' workload when screening for the virus.

Task-specific Fine-tuning via Variational Information Bottleneck for Weakly-supervised Pathology Whole Slide Image Classification

While Multiple Instance Learning (MIL) has shown promising results in digital Pathology Whole Slide Image (WSI) classification, such a paradigm still faces performance and generalization problems due to challenges in high computational costs on Gigapixel WSIs and limited sample size for model training. To deal with the computation problem, most MIL methods utilize a frozen pretrained model from ImageNet to obtain representations first. This process may lose essential information owing to the large domain gap and hinder the generalization of model due to the lack of image-level training-time augmentations. Though Self-supervised Learning (SSL) proposes viable representation learning schemes, the improvement of the downstream task still needs to be further explored in the conversion from the task-agnostic features of SSL to the task-specifics under the partial label supervised learning. To alleviate the dilemma of computation cost and performance, we propose an efficient WSI fine-tuning framework motivated by the Information Bottleneck theory. The theory enables the framework to find the minimal sufficient statistics of WSI, thus supporting us to fine-tune the backbone into a task-specific representation only depending on WSI-level weak labels. The WSI-MIL problem is further analyzed to theoretically deduce our fine-tuning method. Our framework is evaluated on five pathology WSI datasets on various WSI heads. The experimental results of our fine-tuned representations show significant improvements in both accuracy and generalization compared with previous works. Source code will be available at https://github.com/invoker-LL/WSI-finetuning.

Deformable Proposal-Aware P2PNet: A Universal Network for Cell Recognition under Point Supervision

Point-based cell recognition, which aims to localize and classify cells present in a pathology image, is a fundamental task in digital pathology image analysis. The recently developed point-to-point network (P2PNet) has achieved unprecedented cell recognition accuracy and efficiency compared to methods that rely on intermediate density map representations. However, P2PNet could not leverage multi-scale information since it can only decode a single feature map. Moreover, the distribution of predefined point proposals, which is determined by data properties, restricts the resolution of the feature map to decode, i.e., the encoder design. To lift these limitations, we propose a variant of P2PNet named deformable proposal-aware P2PNet (DPA-P2PNet) in this study. The proposed method uses coordinates of point proposals to directly extract multi-scale region-of-interest (ROI) features for feature enhancement. Such a design also opens up possibilities to exploit dynamic distributions of proposals. We further devise a deformation module to improve the proposal quality. Extensive experiments on four datasets with various staining styles demonstrate that DPA-P2PNet outperforms the state-of-the-art methods on point-based cell recognition, which reveals the high potentiality in assisting pathologist assessments.

On-Demand Communication for Asynchronous Multi-Agent Bandits

This paper studies a cooperative multi-agent multi-armed stochastic bandit problem where agents operate asynchronously -- agent pull times and rates are unknown, irregular, and heterogeneous -- and face the same instance of a K-armed bandit problem. Agents can share reward information to speed up the learning process at additional communication costs. We propose ODC, an on-demand communication protocol that tailors the communication of each pair of agents based on their empirical pull times. ODC is efficient when the pull times of agents are highly heterogeneous, and its communication complexity depends on the empirical pull times of agents. ODC is a generic protocol that can be integrated into most cooperative bandit algorithms without degrading their performance. We then incorporate ODC into the natural extensions of UCB and AAE algorithms and propose two communication-efficient cooperative algorithms. Our analysis shows that both algorithms are near-optimal in regret.

Semantic Data Sourcing for 6G Edge Intelligence

As a new function of 6G networks, edge intelligence refers to the ubiquitous deployment of machine learning and artificial intelligence (AI) algorithms at the network edge to empower many emerging applications ranging from sensing to auto-pilot. To support relevant use cases, including sensing, edge learning, and edge inference, all require transmission of high-dimensional data or AI models over the air. To overcome the bottleneck, we propose a novel framework of SEMantic DAta Sourcing (SEMDAS) for locating semantically matched data sources to efficiently enable edge-intelligence operations. The comprehensive framework comprises new architecture, protocol, semantic matching techniques, and design principles for task-oriented wireless techniques. As the key component of SEMDAS, we discuss a set of machine learning based semantic matching techniques targeting different edge-intelligence use cases. Moreover, for designing task-oriented wireless techniques, we discuss different tradeoffs in SEMDAS systems, propose the new concept of joint semantics-and-channel matching, and point to a number of research opportunities. The SEMDAS framework not only overcomes the said communication bottleneck but also addresses other networking issues including long-distance transmission, sparse connectivity, high-speed mobility, link disruptions, and security. In addition, experimental results using a real dataset are presented to demonstrate the performance gain of SEMDAS.

Biomedical image analysis competitions: The state of current participation practice

The number of international benchmarking competitions is steadily increasing in various fields of machine learning (ML) research and practice. So far, however, little is known about the common practice as well as bottlenecks faced by the community in tackling the research questions posed. To shed light on the status quo of algorithm development in the specific field of biomedical imaging analysis, we designed an international survey that was issued to all participants of challenges conducted in conjunction with the IEEE ISBI 2021 and MICCAI 2021 conferences (80 competitions in total). The survey covered participants' expertise and working environments, their chosen strategies, as well as algorithm characteristics. A median of 72% challenge participants took part in the survey. According to our results, knowledge exchange was the primary incentive (70%) for participation, while the reception of prize money played only a minor role (16%). While a median of 80 working hours was spent on method development, a large portion of participants stated that they did not have enough time for method development (32%). 25% perceived the infrastructure to be a bottleneck. Overall, 94% of all solutions were deep learning-based. Of these, 84% were based on standard architectures. 43% of the respondents reported that the data samples (e.g., images) were too large to be processed at once. This was most commonly addressed by patch-based training (69%), downsampling (37%), and solving 3D analysis tasks as a series of 2D tasks. K-fold cross-validation on the training set was performed by only 37% of the participants and only 50% of the participants performed ensembling based on multiple identical models (61%) or heterogeneous models (39%). 48% of the respondents applied postprocessing steps.