Decision Transformer vs. Decision Mamba: Analysing the Complexity of Sequential Decision Making in Atari Games

This work analyses the disparity in performance between Decision Transformer (DT) and Decision Mamba (DM) in sequence modelling reinforcement learning tasks for different Atari games. The study first observed that DM generally outperformed DT in the games Breakout and Qbert, while DT performed better in more complicated games, such as Hero and Kung Fu Master. To understand these differences, we expanded the number of games to 12 and performed a comprehensive analysis of game characteristics, including action space complexity, visual complexity, average trajectory length, and average steps to the first non-zero reward. In order to further analyse the key factors that impact the disparity in performance between DT and DM, we employ various approaches, including quantifying visual complexity, random forest regression, correlation analysis, and action space simplification strategies. The results indicate that the performance gap between DT and DM is affected by the complex interaction of multiple factors, with the complexity of the action space and visual complexity (particularly evaluated by compression ratio) being the primary determining factors. DM performs well in environments with simple action and visual elements, while DT shows an advantage in games with higher action and visual complexity. Our findings contribute to a deeper understanding of how the game characteristics affect the performance difference in sequential modelling reinforcement learning, potentially guiding the development of future model design and applications for diverse and complex environments.

SAFE: Slow and Fast Parameter-Efficient Tuning for Continual Learning with Pre-Trained Models

Continual learning aims to incrementally acquire new concepts in data streams while resisting forgetting previous knowledge. With the rise of powerful pre-trained models (PTMs), there is a growing interest in training incremental learning systems using these foundation models, rather than learning from scratch. Existing works often view PTMs as a strong initial point and directly apply parameter-efficient tuning (PET) in the first session for adapting to downstream tasks. In the following sessions, most methods freeze model parameters for tackling forgetting issues. However, applying PET directly to downstream data cannot fully explore the inherent knowledge in PTMs. Additionally, freezing the parameters in incremental sessions hinders models' plasticity to novel concepts not covered in the first session. To solve the above issues, we propose a Slow And Fast parameter-Efficient tuning (SAFE) framework. In particular, to inherit general knowledge from foundation models, we include a transfer loss function by measuring the correlation between the PTM and the PET-applied model. After calibrating in the first session, the slow efficient tuning parameters can capture more informative features, improving generalization to incoming classes. Moreover, to further incorporate novel concepts, we strike a balance between stability and plasticity by fixing slow efficient tuning parameters and continuously updating the fast ones. Specifically, a cross-classification loss with feature alignment is proposed to circumvent catastrophic forgetting. During inference, we introduce an entropy-based aggregation strategy to dynamically utilize the complementarity in the slow and fast learners. Extensive experiments on seven benchmark datasets verify the effectiveness of our method by significantly surpassing the state-of-the-art.

TAS: Distilling Arbitrary Teacher and Student via a Hybrid Assistant

Most knowledge distillation (KD) methodologies predominantly focus on teacher-student pairs with similar architectures, such as both being convolutional neural networks (CNNs). However, the potential and flexibility of KD can be greatly improved by expanding it to novel Cross-Architecture KD (CAKD), where the knowledge of homogeneous and heterogeneous teachers can be transferred flexibly to a given student. The primary challenge in CAKD lies in the substantial feature gaps between heterogeneous models, originating from the distinction of their inherent inductive biases and module functions. To this end, we introduce an assistant model as a bridge to facilitate smooth feature knowledge transfer between heterogeneous teachers and students. More importantly, within our proposed design principle, the assistant model combines the advantages of cross-architecture inductive biases and module functions by merging convolution and attention modules derived from both student and teacher module functions. Furthermore, we observe that heterogeneous features exhibit diverse spatial distributions in CAKD, hindering the effectiveness of conventional pixel-wise mean squared error (MSE) loss. Therefore, we leverage a spatial-agnostic InfoNCE loss to align features after spatial smoothing, thereby improving the feature alignments in CAKD. Our proposed method is evaluated across some homogeneous model pairs and arbitrary heterogeneous combinations of CNNs, ViTs, and MLPs, achieving state-of-the-art performance for distilled models with a maximum gain of 11.47% on CIFAR-100 and 3.67% on ImageNet-1K. Our code and models will be released.

End-to-end Multi-source Visual Prompt Tuning for Survival Analysis in Whole Slide Images

Survival analysis using pathology images poses a considerable challenge, as it requires the localization of relevant information from the multitude of tiles within whole slide images (WSIs). Current methods typically resort to a two-stage approach, where a pre-trained network extracts features from tiles, which are then used by survival models. This process, however, does not optimize the survival models in an end-to-end manner, and the pre-extracted features may not be ideally suited for survival prediction. To address this limitation, we present a novel end-to-end Visual Prompt Tuning framework for survival analysis, named VPTSurv. VPTSurv refines feature embeddings through an efficient encoder-decoder framework. The encoder remains fixed while the framework introduces tunable visual prompts and adaptors, thus permitting end-to-end training specifically for survival prediction by optimizing only the lightweight adaptors and the decoder. Moreover, the versatile VPTSurv framework accommodates multi-source information as prompts, thereby enriching the survival model. VPTSurv achieves substantial increases of 8.7% and 12.5% in the C-index on two immunohistochemical pathology image datasets. These significant improvements highlight the transformative potential of the end-to-end VPT framework over traditional two-stage methods.

LIDIA: Precise Liver Tumor Diagnosis on Multi-Phase Contrast-Enhanced CT via Iterative Fusion and Asymmetric Contrastive Learning

The early detection and precise diagnosis of liver tumors are tasks of critical clinical value, yet they pose significant challenges due to the high heterogeneity and variability of liver tumors. In this work, a precise LIver tumor DIAgnosis network on multi-phase contrast-enhance CT, named LIDIA, is proposed for real-world scenario. To fully utilize all available phases in contrast-enhanced CT, LIDIA first employs the iterative fusion module to aggregate variable numbers of image phases, thereby capturing the features of lesions at different phases for better tumor diagnosis. To effectively mitigate the high heterogeneity problem of liver tumors, LIDIA incorporates asymmetric contrastive learning to enhance the discriminability between different classes. To evaluate our method, we constructed a large-scale dataset comprising 1,921 patients and 8,138 lesions. LIDIA has achieved an average AUC of 93.6% across eight different types of lesions, demonstrating its effectiveness. Besides, LIDIA also demonstrated strong generalizability with an average AUC of 89.3% when tested on an external cohort of 828 patients.

Improved Esophageal Varices Assessment from Non-Contrast CT Scans

Esophageal varices (EV), a serious health concern resulting from portal hypertension, are traditionally diagnosed through invasive endoscopic procedures. Despite non-contrast computed tomography (NC-CT) imaging being a less expensive and non-invasive imaging modality, it has yet to gain full acceptance as a primary clinical diagnostic tool for EV evaluation. To overcome existing diagnostic challenges, we present the Multi-Organ-cOhesion-Network (MOON), a novel framework enhancing the analysis of critical organ features in NC-CT scans for effective assessment of EV. Drawing inspiration from the thorough assessment practices of radiologists, MOON establishes a cohesive multiorgan analysis model that unifies the imaging features of the related organs of EV, namely esophagus, liver, and spleen. This integration significantly increases the diagnostic accuracy for EV. We have compiled an extensive NC-CT dataset of 1,255 patients diagnosed with EV, spanning three grades of severity. Each case is corroborated by endoscopic diagnostic results. The efficacy of MOON has been substantiated through a validation process involving multi-fold cross-validation on 1,010 cases and an independent test on 245 cases, exhibiting superior diagnostic performance compared to methods focusing solely on the esophagus (for classifying severe grade: AUC of 0.864 versus 0.803, and for moderate to severe grades: AUC of 0.832 versus 0.793). To our knowledge, MOON is the first work to incorporate a synchronized multi-organ NC-CT analysis for EV assessment, providing a more acceptable and minimally invasive alternative for patients compared to traditional endoscopy.

Dual Relation Mining Network for Zero-Shot Learning

Zero-shot learning (ZSL) aims to recognize novel classes through transferring shared semantic knowledge (e.g., attributes) from seen classes to unseen classes. Recently, attention-based methods have exhibited significant progress which align visual features and attributes via a spatial attention mechanism. However, these methods only explore visual-semantic relationship in the spatial dimension, which can lead to classification ambiguity when different attributes share similar attention regions, and semantic relationship between attributes is rarely discussed. To alleviate the above problems, we propose a Dual Relation Mining Network (DRMN) to enable more effective visual-semantic interactions and learn semantic relationship among attributes for knowledge transfer. Specifically, we introduce a Dual Attention Block (DAB) for visual-semantic relationship mining, which enriches visual information by multi-level feature fusion and conducts spatial attention for visual to semantic embedding. Moreover, an attribute-guided channel attention is utilized to decouple entangled semantic features. For semantic relationship modeling, we utilize a Semantic Interaction Transformer (SIT) to enhance the generalization of attribute representations among images. Additionally, a global classification branch is introduced as a complement to human-defined semantic attributes, and we then combine the results with attribute-based classification. Extensive experiments demonstrate that the proposed DRMN leads to new state-of-the-art performances on three standard ZSL benchmarks, i.e., CUB, SUN, and AwA2.

Artificial intelligence for context-aware visual change detection in software test automation

Automated software testing is integral to the software development process, streamlining workflows and ensuring product reliability. Visual testing within this context, especially concerning user interface (UI) and user experience (UX) validation, stands as one of crucial determinants of overall software quality. Nevertheless, conventional methods like pixel-wise comparison and region-based visual change detection fall short in capturing contextual similarities, nuanced alterations, and understanding the spatial relationships between UI elements. In this paper, we introduce a novel graph-based method for visual change detection in software test automation. Leveraging a machine learning model, our method accurately identifies UI controls from software screenshots and constructs a graph representing contextual and spatial relationships between the controls. This information is then used to find correspondence between UI controls within screenshots of different versions of a software. The resulting graph encapsulates the intricate layout of the UI and underlying contextual relations, providing a holistic and context-aware model. This model is finally used to detect and highlight visual regressions in the UI. Comprehensive experiments on different datasets showed that our change detector can accurately detect visual software changes in various simple and complex test scenarios. Moreover, it outperformed pixel-wise comparison and region-based baselines by a large margin in more complex testing scenarios. This work not only contributes to the advancement of visual change detection but also holds practical implications, offering a robust solution for real-world software test automation challenges, enhancing reliability, and ensuring the seamless evolution of software interfaces.

CT-GLIP: 3D Grounded Language-Image Pretraining with CT Scans and Radiology Reports for Full-Body Scenarios

Medical Vision-Language Pretraining (Med-VLP) establishes a connection between visual content from medical images and the relevant textual descriptions. Existing Med-VLP methods primarily focus on 2D images depicting a single body part, notably chest X-rays. In this paper, we extend the scope of Med-VLP to encompass 3D images, specifically targeting full-body scenarios, by using a multimodal dataset of CT images and reports. Compared with the 2D counterpart, 3D VLP is required to effectively capture essential semantics from significantly sparser representation in 3D imaging. In this paper, we introduce CT-GLIP (Grounded Language-Image Pretraining with CT scans), a novel method that constructs organ-level image-text pairs to enhance multimodal contrastive learning, aligning grounded visual features with precise diagnostic text. Additionally, we developed an abnormality dictionary to augment contrastive learning with diverse contrastive pairs. Our method, trained on a multimodal CT dataset comprising 44,011 organ-level vision-text pairs from 17,702 patients across 104 organs, demonstrates it can identify organs and abnormalities in a zero-shot manner using natural languages. The performance of CT-GLIP is validated on a separate test set of 1,130 patients, focusing on the 16 most frequent abnormalities across 7 organs. The experimental results show our model's superior performance over the standard CLIP framework across zero-shot and fine-tuning scenarios, using both CNN and ViT architectures.

Anchor-based Robust Finetuning of Vision-Language Models

We aim at finetuning a vision-language model without hurting its out-of-distribution (OOD) generalization. We address two types of OOD generalization, i.e., i) domain shift such as natural to sketch images, and ii) zero-shot capability to recognize the category that was not contained in the finetune data. Arguably, the diminished OOD generalization after finetuning stems from the excessively simplified finetuning target, which only provides the class information, such as ``a photo of a [CLASS]''. This is distinct from the process in that CLIP was pretrained, where there is abundant text supervision with rich semantic information. Therefore, we propose to compensate for the finetune process using auxiliary supervision with rich semantic information, which acts as anchors to preserve the OOD generalization. Specifically, two types of anchors are elaborated in our method, including i) text-compensated anchor which uses the images from the finetune set but enriches the text supervision from a pretrained captioner, ii) image-text-pair anchor which is retrieved from the dataset similar to pretraining data of CLIP according to the downstream task, associating with the original CLIP text with rich semantics. Those anchors are utilized as auxiliary semantic information to maintain the original feature space of CLIP, thereby preserving the OOD generalization capabilities. Comprehensive experiments demonstrate that our method achieves in-distribution performance akin to conventional finetuning while attaining new state-of-the-art results on domain shift and zero-shot learning benchmarks.