Search

Reducing computational costs is an important issue for development of embedded systems. Binary-weight Neural Networks (BNNs), in which weights are binarized and activations are quantized, are employed to reduce computational costs of various kinds of applications. In this paper, a design methodology of hardware architecture for inference engines is proposed to handle modern BNNs with two operation modes. Multiply-Accumulate (MAC) operations can be simplified by replacing multiply operations with bitwise operations. The proposed method can effectively reduce the gate count of inference engines by removing a part of computational costs from the hardware system. The architecture of MAC operations can calculate the inference results of BNNs efficiently with only 52% of hardware costs compared with the related works. To show that the inference engine can handle practical applications, two lightweight networks which combine the backbones of SegNeXt and the decoder of SparseInst for instance segmentation are also proposed. The output results of the lightweight networks are computed using only bitwise operations and add operations. The proposed inference engine has lower hardware costs than related works. The experimental results show that the proposed inference engine can handle the proposed instance-segmentation networks and achieves higher accuracy than YOLACT on the "Person" category although the model size is 77.7$\times$ smaller compared with YOLACT., Comment: Camera-ready version for CVPR 2024 workshop (Embedded Vision Workshop)

U-Net models with encoder, decoder, and skip-connections components have demonstrated effectiveness in a variety of vision tasks. The skip-connections transmit fine-grained information from the encoder to the decoder. It is necessary to maintain the feature maps used by the skip-connections in memory before the decoding stage. Therefore, they are not friendly to devices with limited resource. In this paper, we propose a universal method and architecture to reduce the memory consumption and meanwhile generate enhanced feature maps to improve network performance. To this end, we design a simple but effective Multi-Scale Information Aggregation Module (MSIAM) in the encoder and an Information Enhancement Module (IEM) in the decoder. The MSIAM aggregates multi-scale feature maps into single-scale with less memory. After that, the aggregated feature maps can be expanded and enhanced to multi-scale feature maps by the IEM. By applying the proposed method on NAFNet, a SOTA model in the field of image restoration, we design a memory-efficient and feature-enhanced network architecture, UNet--. The memory demand by the skip-connections in the UNet-- is reduced by 93.3%, while the performance is improved compared to NAFNet. Furthermore, we show that our proposed method can be generalized to multiple visual tasks, with consistent improvements in both memory consumption and network accuracy compared to the existing efficient architectures., Comment: 17 pages, 7 figures, accepted by ACCV2024

Background: The worst outcomes of diabetic retinopathy (DR) can be prevented by implementing DR screening programs assisted by AI. At the University Hospital of Navarre (HUN), Spain, general practitioners (GPs) grade fundus images in an ongoing DR screening program, referring to a second screening level (ophthalmologist) target patients. Methods: After collecting their requirements, HUN decided to develop a custom AI tool, called NaIA-RD, to assist their GPs in DR screening. This paper introduces NaIA-RD, details its implementation, and highlights its unique combination of DR and retinal image quality grading in a single system. Its impact is measured in an unprecedented before-and-after study that compares 19,828 patients screened before NaIA-RD's implementation and 22,962 patients screened after. Results: NaIA-RD influenced the screening criteria of 3/4 GPs, increasing their sensitivity. Agreement between NaIA-RD and the GPs was high for non-referral proposals (94.6% or more), but lower and variable (from 23.4\% to 86.6%) for referral proposals. An ophthalmologist discarded a NaIA-RD error in most of contradicted referral proposals by labeling the 93% of a sample of them as referable. In an autonomous setup, NaIA-RD would have reduced the study visualization workload by 4.27 times without missing a single case of sight-threatening DR referred by a GP. Conclusion: DR screening was more effective when supported by NaIA-RD, which could be safely used to autonomously perform the first level of screening. This shows how AI devices, when seamlessly integrated into clinical workflows, can help improve clinical pathways in the long term.

Recent studies on online reinforcement learning (RL) have demonstrated the advantages of learning multiple behaviors from a single task, as in the case of few-shot adaptation to a new environment. Although this approach is expected to yield similar benefits in offline RL, appropriate methods for learning multiple solutions have not been fully investigated in previous studies. In this study, we therefore addressed the problem of finding multiple solutions from a single task in offline RL. We propose algorithms that can learn multiple solutions in offline RL, and empirically investigate their performance. Our experimental results show that the proposed algorithm learns multiple qualitatively and quantitatively distinctive solutions in offline RL., Comment: ICML 2024, 21 pages

In offline reinforcement learning, in-sample learning methods have been widely used to prevent performance degradation caused by evaluating out-of-distribution actions from the dataset. Extreme Q-learning (XQL) employs a loss function based on the assumption that Bellman error follows a Gumbel distribution, enabling it to model the soft optimal value function in an in-sample manner. It has demonstrated strong performance in both offline and online reinforcement learning settings. However, issues remain, such as the instability caused by the exponential term in the loss function and the risk of the error distribution deviating from the Gumbel distribution. Therefore, we propose Maclaurin Expanded Extreme Q-learning to enhance stability. In this method, applying Maclaurin expansion to the loss function in XQL enhances stability against large errors. This approach involves adjusting the modeled value function between the value function under the behavior policy and the soft optimal value function, thus achieving a trade-off between stability and optimality depending on the order of expansion. It also enables adjustment of the error distribution assumption from a normal distribution to a Gumbel distribution. Our method significantly stabilizes learning in online RL tasks from DM Control, where XQL was previously unstable. Additionally, it improves performance in several offline RL tasks from D4RL., Comment: Accepted at RLC 2024: The first Reinforcement Learning Conference

Current Reinforcement Learning (RL) is often limited by the large amount of data needed to learn a successful policy. Offline RL aims to solve this issue by using transitions collected by a different behavior policy. We address a novel Offline RL problem setting in which, while collecting the dataset, the transition and reward functions gradually change between episodes but stay constant within each episode. We propose a method based on Contrastive Predictive Coding that identifies this non-stationarity in the offline dataset, accounts for it when training a policy, and predicts it during evaluation. We analyze our proposed method and show that it performs well in simple continuous control tasks and challenging, high-dimensional locomotion tasks. We show that our method often achieves the oracle performance and performs better than baselines., Comment: Accepted for Reinforcement Learning Conference (RLC) 2024

The Photon Detection System (PDS) of the first DUNE far detector (FD1) is composed of 6000 photon detection units, named X-ARAPUCA. The detection of the prompt light pulse generated by the particle energy release in liquid argon (LAr) will complement and boost the DUNE Liquid Argon Time Projection Chamber (LArTPC). It will improve the non-beam events tagging and enable at low energies the trigger and the calorimetry of the supernova neutrinos. The X-ARAPUCA unit is an assembly of several components. Its Photon Detection Efficiency (PDE) depends both on the design of the assembly, on the grade of the individual components and finally on their coupling. The X-ARAPUCA PDE is one of the leading parameters for the Photon Detection System sensitivity, that in turn determines the sensitivity of the DUNE for the detection of core-collapse supernova within the galaxy and for nucleon decay searches. In this work we present the final assessment of the absolute PDE of the FD1 X-ARAPUCA baseline design, measured in two laboratories with independent methods and setups. One hundred sixty units of these X-ARAPUCA devices have been deployed in the NP04 facility at the CERN Neutrino Platform, the 1:20 scale FD1 prototype, and will be operated during the year 2024. The assessed value of the PDE is a key parameter both in the NP04 and in the DUNE analysis and reconstruction studies.

In deep reinforcement learning, estimating the value function to evaluate the quality of states and actions is essential. The value function is often trained using the least squares method, which implicitly assumes a Gaussian error distribution. However, a recent study suggested that the error distribution for training the value function is often skewed because of the properties of the Bellman operator, and violates the implicit assumption of normal error distribution in the least squares method. To address this, we proposed a method called Symmetric Q-learning, in which the synthetic noise generated from a zero-mean distribution is added to the target values to generate a Gaussian error distribution. We evaluated the proposed method on continuous control benchmark tasks in MuJoCo. It improved the sample efficiency of a state-of-the-art reinforcement learning method by reducing the skewness of the error distribution., Comment: Accepted at AAAI 2024: The 38th Annual AAAI Conference on Artificial Intelligence (Main Tech Track)

This paper describes a procedure for the validation of alpha-particle sources (exempt unsealed sources) to be used in experimental setups with liquefied gases at cryogenic temperatures (down to -196 C) and high vacuum. These setups are of interest for the development and characterization of neutrino and dark matter detectors based on liquid argon, among others. Due to the high purity requirements, the sources have to withstand high vacuum and cryogenic temperatures for extended periods. The validation procedure has been applied to 241Am sources produced by electrodeposition., Comment: 5 pages

Autonomous assistance of people with motor impairments is one of the most promising applications of autonomous robotic systems. Recent studies have reported encouraging results using deep reinforcement learning (RL) in the healthcare domain. Previous studies showed that assistive tasks can be formulated as multi-agent RL, wherein there are two agents: a caregiver and a care-receiver. However, policies trained in multi-agent RL are often sensitive to the policies of other agents. In such a case, a trained caregiver's policy may not work for different care-receivers. To alleviate this issue, we propose a framework that learns a robust caregiver's policy by training it for diverse care-receiver responses. In our framework, diverse care-receiver responses are autonomously learned through trials and errors. In addition, to robustify the care-giver's policy, we propose a strategy for sampling a care-receiver's response in an adversarial manner during the training. We evaluated the proposed method using tasks in an Assistive Gym. We demonstrate that policies trained with a popular deep RL method are vulnerable to changes in policies of other agents and that the proposed framework improves the robustness against such changes., Comment: 7 pages, accepted for ICRA 2024

Liquid argon time projection chambers (TPC) are widely used in neutrino oscillation and dark matter experiments. Detection of scintillation light in liquid argon TPC's is challenging because of its short wavelength, in the VUV range, and the cryogenic temperatures (~86 K) at which the sensors must operate. Wavelength shifters (WLS) are typically needed to take advantage of the high Photon Detection Efficiency (PDE) in the visible range of most of photondetectors. The Hamamatsu VUV4 S13370--6075CN SiPMs can directly detect VUV light without the use of WLS, which main benefit is an improved PDE at these short wavelengths, but also the visible light from WLS. The manufacturer (Hamamatsu Photonics K.K.) provides a complete characterization of these devices at room temperature; however, previous studies have indicated a decrease of the PDE at cryogenic temperature for VUV light. In this work, we present the measurement of the PDE of VUV4 SiPMs at cryogenic temperature for different wavelengths in the range [270, 570] nm. A dedicated measurement at 127 nm is also shown., Comment: F6a,F9b:clear L145,L154,L233,L226-231:rephrased 5.2TitleChanged L68,L151,L165:typos Additional measurements showed that the V that was being used in the LAr setup was obtained for LN2 temperature.To compare with 4OV we have applied a 3% correction to account for the PDEvsOV dependance. Removed unused references and added some missing acknowledgments

Glioblastoma is one of the most aggressive and deadliest types of brain cancer, with low survival rates compared to other types of cancer. Analysis of Magnetic Resonance Imaging (MRI) scans is one of the most effective methods for the diagnosis and treatment of brain cancers such as glioblastoma. Accurate tumor segmentation in MRI images is often required for treatment planning and risk assessment of treatment methods. Here, we propose a novel pipeline, Brain Radiology Aided by Intelligent Neural NETworks (BRAINNET), which leverages MaskFormer, a vision transformer model, and generates robust tumor segmentation maks. We use an ensemble of nine predictions from three models separately trained on each of the three orthogonal 2D slice directions (axial, sagittal, and coronal) of a 3D brain MRI volume. We train and test our models on the publicly available UPenn-GBM dataset, consisting of 3D multi-parametric MRI (mpMRI) scans from 611 subjects. Using Dice coefficient (DC) and 95% Hausdorff distance (HD) for evaluation, our models achieved state-of-the-art results in segmenting all three different tumor regions -- tumor core (DC = 0.894, HD = 2.308), whole tumor (DC = 0.891, HD = 3.552), and enhancing tumor (DC = 0.812, HD = 1.608).

Large, high-capacity models trained on diverse datasets have shown remarkable successes on efficiently tackling downstream applications. In domains from NLP to Computer Vision, this has led to a consolidation of pretrained models, with general pretrained backbones serving as a starting point for many applications. Can such a consolidation happen in robotics? Conventionally, robotic learning methods train a separate model for every application, every robot, and even every environment. Can we instead train generalist X-robot policy that can be adapted efficiently to new robots, tasks, and environments? In this paper, we provide datasets in standardized data formats and models to make it possible to explore this possibility in the context of robotic manipulation, alongside experimental results that provide an example of effective X-robot policies. We assemble a dataset from 22 different robots collected through a collaboration between 21 institutions, demonstrating 527 skills (160266 tasks). We show that a high-capacity model trained on this data, which we call RT-X, exhibits positive transfer and improves the capabilities of multiple robots by leveraging experience from other platforms. More details can be found on the project website https://robotics-transformer-x.github.io., Comment: Project website: https://robotics-transformer-x.github.io