Hyper: Hyperparameter Robust Efficient Exploration in Reinforcement Learning

The exploration \& exploitation dilemma poses significant challenges in reinforcement learning (RL). Recently, curiosity-based exploration methods achieved great success in tackling hard-exploration problems. However, they necessitate extensive hyperparameter tuning on different environments, which heavily limits the applicability and accessibility of this line of methods. In this paper, we characterize this problem via analysis of the agent behavior, concluding the fundamental difficulty of choosing a proper hyperparameter. We then identify the difficulty and the instability of the optimization when the agent learns with curiosity. We propose our method, hyperparameter robust exploration (\textbf{Hyper}), which extensively mitigates the problem by effectively regularizing the visitation of the exploration and decoupling the exploitation to ensure stable training. We theoretically justify that \textbf{Hyper} is provably efficient under function approximation setting and empirically demonstrate its appealing performance and robustness in various environments.

A Survey on Deep Clustering: From the Prior Perspective

Facilitated by the powerful feature extraction ability of neural networks, deep clustering has achieved great success in analyzing high-dimensional and complex real-world data. The performance of deep clustering methods is affected by various factors such as network structures and learning objectives. However, as pointed out in this survey, the essence of deep clustering lies in the incorporation and utilization of prior knowledge, which is largely ignored by existing works. From pioneering deep clustering methods based on data structure assumptions to recent contrastive clustering methods based on data augmentation invariances, the development of deep clustering intrinsically corresponds to the evolution of prior knowledge. In this survey, we provide a comprehensive review of deep clustering methods by categorizing them into six types of prior knowledge. We find that in general the prior innovation follows two trends, namely, i) from mining to constructing, and ii) from internal to external. Besides, we provide a benchmark on five widely-used datasets and analyze the performance of methods with diverse priors. By providing a novel prior knowledge perspective, we hope this survey could provide some novel insights and inspire future research in the deep clustering community.

The RoboDrive Challenge: Drive Anytime Anywhere in Any Condition

In the realm of autonomous driving, robust perception under out-of-distribution conditions is paramount for the safe deployment of vehicles. Challenges such as adverse weather, sensor malfunctions, and environmental unpredictability can severely impact the performance of autonomous systems. The 2024 RoboDrive Challenge was crafted to propel the development of driving perception technologies that can withstand and adapt to these real-world variabilities. Focusing on four pivotal tasks -- BEV detection, map segmentation, semantic occupancy prediction, and multi-view depth estimation -- the competition laid down a gauntlet to innovate and enhance system resilience against typical and atypical disturbances. This year's challenge consisted of five distinct tracks and attracted 140 registered teams from 93 institutes across 11 countries, resulting in nearly one thousand submissions evaluated through our servers. The competition culminated in 15 top-performing solutions, which introduced a range of innovative approaches including advanced data augmentation, multi-sensor fusion, self-supervised learning for error correction, and new algorithmic strategies to enhance sensor robustness. These contributions significantly advanced the state of the art, particularly in handling sensor inconsistencies and environmental variability. Participants, through collaborative efforts, pushed the boundaries of current technologies, showcasing their potential in real-world scenarios. Extensive evaluations and analyses provided insights into the effectiveness of these solutions, highlighting key trends and successful strategies for improving the resilience of driving perception systems. This challenge has set a new benchmark in the field, providing a rich repository of techniques expected to guide future research in this field.

An Empirical Study of Parameter Efficient Fine-tuning on Vision-Language Pre-train Model

Recent studies applied Parameter Efficient Fine-Tuning techniques (PEFTs) to efficiently narrow the performance gap between pre-training and downstream. There are two important factors for various PEFTs, namely, the accessible data size and fine-tunable parameter size. A natural expectation for PEFTs is that the performance of various PEFTs is positively related to the data size and fine-tunable parameter size. However, according to the evaluation of five PEFTs on two downstream vision-language (VL) tasks, we find that such an intuition holds only if the downstream data and task are not consistent with pre-training. For downstream fine-tuning consistent with pre-training, data size no longer affects the performance, while the influence of fine-tunable parameter size is not monotonous. We believe such an observation could guide the choice of training strategy for various PEFTs.

Prognostic Covariate Adjustment for Logistic Regression in Randomized Controlled Trials

Randomized controlled trials (RCTs) with binary primary endpoints introduce novel challenges for inferring the causal effects of treatments. The most significant challenge is non-collapsibility, in which the conditional odds ratio estimand under covariate adjustment differs from the unconditional estimand in the logistic regression analysis of RCT data. This issue gives rise to apparent paradoxes, such as the variance of the estimator for the conditional odds ratio from a covariate-adjusted model being greater than the variance of the estimator from the unadjusted model. We address this challenge in the context of adjustment based on predictions of control outcomes from generative artificial intelligence (AI) algorithms, which are referred to as prognostic scores. We demonstrate that prognostic score adjustment in logistic regression increases the power of the Wald test for the conditional odds ratio under a fixed sample size, or alternatively reduces the necessary sample size to achieve a desired power, compared to the unadjusted analysis. We derive formulae for prospective calculations of the power gain and sample size reduction that can result from adjustment for the prognostic score. Furthermore, we utilize g-computation to expand the scope of prognostic score adjustment to inferences on the marginal risk difference, relative risk, and odds ratio estimands. We demonstrate the validity of our formulae via extensive simulation studies that encompass different types of logistic regression model specifications. Our simulation studies also indicate how prognostic score adjustment can reduce the variance of g-computation estimators for the marginal estimands while maintaining frequentist properties such as asymptotic unbiasedness and Type I error rate control. Our methodology can ultimately enable more definitive and conclusive analyses for RCTs with binary primary endpoints.

Achieving Near-Optimal Regret for Bandit Algorithms with Uniform Last-Iterate Guarantee

Existing performance measures for bandit algorithms such as regret, PAC bounds, or uniform-PAC (Dann et al., 2017), typically evaluate the cumulative performance, while allowing the play of an arbitrarily bad arm at any finite time t. Such a behavior can be highly detrimental in high-stakes applications. This paper introduces a stronger performance measure, the uniform last-iterate (ULI) guarantee, capturing both cumulative and instantaneous performance of bandit algorithms. Specifically, ULI characterizes the instantaneous performance since it ensures that the per-round regret of the played arm is bounded by a function, monotonically decreasing w.r.t. (large) round t, preventing revisits to bad arms when sufficient samples are available. We demonstrate that a near-optimal ULI guarantee directly implies near-optimal cumulative performance across aforementioned performance measures. To examine the achievability of ULI in the finite arm setting, we first provide two positive results that some elimination-based algorithms and high-probability adversarial algorithms with stronger analysis or additional designs, can attain near-optimal ULI guarantees. Then, we also provide a negative result, indicating that optimistic algorithms cannot achieve a near-optimal ULI guarantee. Finally, we propose an efficient algorithm for linear bandits with infinitely many arms, which achieves the ULI guarantee, given access to an optimization oracle.

Image Clustering with External Guidance

The core of clustering is incorporating prior knowledge to construct supervision signals. From classic k-means based on data compactness to recent contrastive clustering guided by self-supervision, the evolution of clustering methods intrinsically corresponds to the progression of supervision signals. At present, substantial efforts have been devoted to mining internal supervision signals from data. Nevertheless, the abundant external knowledge such as semantic descriptions, which naturally conduces to clustering, is regrettably overlooked. In this work, we propose leveraging external knowledge as a new supervision signal to guide clustering, even though it seems irrelevant to the given data. To implement and validate our idea, we design an externally guided clustering method (Text-Aided Clustering, TAC), which leverages the textual semantics of WordNet to facilitate image clustering. Specifically, TAC first selects and retrieves WordNet nouns that best distinguish images to enhance the feature discriminability. Then, to improve image clustering performance, TAC collaborates text and image modalities by mutually distilling cross-modal neighborhood information. Experiments demonstrate that TAC achieves state-of-the-art performance on five widely used and three more challenging image clustering benchmarks, including the full ImageNet-1K dataset.

Automated Assessment of Critical View of Safety in Laparoscopic Cholecystectomy

Cholecystectomy (gallbladder removal) is one of the most common procedures in the US, with more than 1.2M procedures annually. Compared with classical open cholecystectomy, laparoscopic cholecystectomy (LC) is associated with significantly shorter recovery period, and hence is the preferred method. However, LC is also associated with an increase in bile duct injuries (BDIs), resulting in significant morbidity and mortality. The primary cause of BDIs from LCs is misidentification of the cystic duct with the bile duct. Critical view of safety (CVS) is the most effective of safety protocols, which is said to be achieved during the surgery if certain criteria are met. However, due to suboptimal understanding and implementation of CVS, the BDI rates have remained stable over the last three decades. In this paper, we develop deep-learning techniques to automate the assessment of CVS in LCs. An innovative aspect of our research is on developing specialized learning techniques by incorporating domain knowledge to compensate for the limited training data available in practice. In particular, our CVS assessment process involves a fusion of two segmentation maps followed by an estimation of a certain region of interest based on anatomical structures close to the gallbladder, and then finally determination of each of the three CVS criteria via rule-based assessment of structural information. We achieved a gain of over 11.8% in mIoU on relevant classes with our two-stream semantic segmentation approach when compared to a single-model baseline, and 1.84% in mIoU with our proposed Sobel loss function when compared to a Transformer-based baseline model. For CVS criteria, we achieved up to 16% improvement and, for the overall CVS assessment, we achieved 5% improvement in balanced accuracy compared to DeepCVS under the same experiment settings.

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.