Retrofit: Continual Learning with Bounded Forgetting for Security Applications

Saliency maps are essential tools for providing visual explanations in deep learning, yet there is a significant lack of consensus on their purpose and alignment with user queries. This ambiguity complicates the evaluation and practical application of explanation methods. The introduction of the Reference-Frame × Granularity (RFxG) taxonomy aims to address this issue by categorizing saliency explanations based on two axes: Reference-Frame and Granularity, highlighting limitations in current evaluation metrics.

The study addresses the challenge of blood coagulation after drawing, which can lead to inaccurate hematological analyses. It presents a deep learning-enhanced framework utilizing Digital Holographic Microscopy (DHM) to evaluate anticoagulant efficacy ex vivo. The method allows for label-free, non-invasive analysis of human blood samples, focusing on cell counting and morphology. The research compares the effects of conventional EDTA and potassium ferric oxalate nanoparticles (KFeOx-NPs) on blood coagulation rates, revealing significant findings.

SimuFreeMark is a proposed watermarking framework designed to enhance image security against editing attacks, particularly in the context of artificial intelligence-generated content (AIGC). Unlike existing methods that depend on noise simulation, SimuFreeMark directly embeds watermarks into the low-frequency components of images, which have shown significant robustness against various attacks. This innovation aims to address the growing need for reliable watermarking solutions in an era of advanced image manipulation techniques.

An AI-assisted Augmented Reality (AR) assembly workflow has been developed, utilizing deep learning-based object recognition to identify assembly components and provide step-by-step instructions. The system displays bounding boxes around components in real-time, indicating their placement, thus eliminating the need for manual searching or sorting. A case study involving the assembly of LEGO sculptures demonstrates the system's feasibility and effectiveness in enhancing the assembly process.

A new study presents a deep learning framework for short-term precipitation prediction in Bengaluru, Mumbai, Delhi, and Kolkata, utilizing a hybrid CNN-ConvLSTM architecture. This model, trained on multi-decadal ERA5 reanalysis data, aims to enhance transparency in weather forecasting. The models achieved varying root mean square error (RMSE) values: 0.21 mm/day for Bengaluru, 0.52 mm/day for Mumbai, 0.48 mm/day for Delhi, and 1.80 mm/day for Kolkata. The approach emphasizes explainable AI to improve understanding of precipitation patterns.

NervePool is a newly proposed pooling layer designed for deep learning applications involving graph-structured data. It focuses on simplicial complexes, which extend traditional graphs by incorporating higher-dimensional simplices. This innovation aims to enhance the modeling of complex relationships while addressing issues such as computational efficiency and overfitting. The pooling layer operates through a learned hierarchical representation, allowing for effective down sampling and information compression.

The article titled 'Bayesian ICA with super-Gaussian Source Priors' discusses advancements in Independent Component Analysis (ICA), a key method in machine learning for feature extraction. The authors introduce a horseshoe-type prior with a latent Polya-Gamma scale mixture representation, enabling scalable algorithms for point estimation and full posterior inference. The study establishes theoretical guarantees for hierarchical Bayesian ICA, including results for the unmixing matrix. Simulation studies indicate that the proposed methods are competitive with existing ICA tools.

The research paper titled 'DRMD: Deep Reinforcement Learning for Malware Detection under Concept Drift' addresses the challenges of malware detection in dynamic environments. It highlights the limitations of traditional classifiers in adapting to evolving threats and proposes a novel approach using deep reinforcement learning (DRL). The DRL agent optimizes sample classification while managing high-risk samples for manual labeling, demonstrating improved performance metrics over standard methods.