New submissions for Mon, 25 Jul 22

[ericbeyer]

Keyword: out of distribution detection

There is no result

Keyword: out-of-distribution detection

There is no result

Keyword: expected calibration error

There is no result

Keyword: overconfident

There is no result

Keyword: overconfidence

There is no result

Keyword: confidence

Learning Deep SDF Maps Online for Robot Navigation and Exploration

• Authors: Gadiel Sznaier Camps, Robert Dyro, Marco Pavone, Mac Schwager
• Subjects: Robotics (cs.RO)
• Arxiv link: https://arxiv.org/abs/2207.10782
• Pdf link: https://arxiv.org/pdf/2207.10782
• Abstract We propose an algorithm to (i) learn online a deep signed distance function (SDF) with a LiDAR-equipped robot to represent the 3D environment geometry, and (ii) plan collision-free trajectories given this deep learned map. Our algorithm takes a stream of incoming LiDAR scans and continually optimizes a neural network to represent the SDF of the environment around its current vicinity. When the SDF network quality saturates, we cache a copy of the network, along with a learned confidence metric, and initialize a new SDF network to continue mapping new regions of the environment. We then concatenate all the cached local SDFs through a confidence-weighted scheme to give a global SDF for planning. For planning, we make use of a sequential convex model predictive control (MPC) algorithm. The MPC planner optimizes a dynamically feasible trajectory for the robot while enforcing no collisions with obstacles mapped in the global SDF. We show that our online mapping algorithm produces higher-quality maps than existing methods for online SDF training. In the WeBots simulator, we further showcase the combined mapper and planner running online -- navigating autonomously and without collisions in an unknown environment.

  Keyword: scaling

  The Dirty Secret of SSDs: Embodied Carbon

• Authors: Swamit Tannu, Prashant J. Nair
• Subjects: Hardware Architecture (cs.AR); Computers and Society (cs.CY)
• Arxiv link: https://arxiv.org/abs/2207.10793
• Pdf link: https://arxiv.org/pdf/2207.10793
• Abstract Scalable Solid-State Drives (SSDs) have revolutionized the way we store and access our data across datacenters and handheld devices. Unfortunately, scaling technology can have a significant environmental impact. Across the globe, most semiconductor manufacturing use electricity that is generated from coal and natural gas. For instance, manufacturing a Gigabyte of Flash emits 0.16 Kg CO$_2$ and is a significant fraction of the total carbon emission in the system. We estimate that manufacturing storage devices has resulted in 20 million metric tonnes of CO$_2$ emissions in 2021 alone. To better understand this concern, this paper compares the sustainability trade-offs between Hard Disk Drives (HDDs) and SSDs and recommends methodologies to estimate the embodied carbon costs of the storage system. In this paper, we outline four possible strategies to make storage systems sustainable. First, this paper recommends directions that help select the right medium of storage (SSD vs HDD). Second, this paper proposes lifetime extension techniques for SSDs. Third, this paper advocates for effective and efficient recycling and reuse of high-density multi-level cell-based SSDs. Fourth, specifically for hand-held devices, this paper recommends leveraging elasticity in cloud storage.

  Scale dependant layer for self-supervised nuclei encoding

• Authors: Peter Naylor, Yao-Hung Hubert Tsai, Marick Laé, Makoto Yamada
• Subjects: Computer Vision and Pattern Recognition (cs.CV)
• Arxiv link: https://arxiv.org/abs/2207.10950
• Pdf link: https://arxiv.org/pdf/2207.10950
• Abstract Recent developments in self-supervised learning give us the possibility to further reduce human intervention in multi-step pipelines where the focus evolves around particular objects of interest. In the present paper, the focus lays in the nuclei in histopathology images. In particular we aim at extracting cellular information in an unsupervised manner for a downstream task. As nuclei present themselves in a variety of sizes, we propose a new Scale-dependant convolutional layer to bypass scaling issues when resizing nuclei. On three nuclei datasets, we benchmark the following methods: handcrafted, pre-trained ResNet, supervised ResNet and self-supervised features. We show that the proposed convolution layer boosts performance and that this layer combined with Barlows-Twins allows for better nuclei encoding compared to the supervised paradigm in the low sample setting and outperforms all other proposed unsupervised methods. In addition, we extend the existing TNBC dataset to incorporate nuclei class annotation in order to enrich and publicly release a small sample setting dataset for nuclei segmentation and classification.

  Decentralized scheduling through an adaptive, trading-based multi-agent system

• Authors: Michael Kölle, Lennart Rietdorf, Kyrill Schmid
• Subjects: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
• Arxiv link: https://arxiv.org/abs/2207.11172
• Pdf link: https://arxiv.org/pdf/2207.11172
• Abstract In multi-agent reinforcement learning systems, the actions of one agent can have a negative impact on the rewards of other agents. One way to combat this problem is to let agents trade their rewards amongst each other. Motivated by this, this work applies a trading approach to a simulated scheduling environment, where the agents are responsible for the assignment of incoming jobs to compute cores. In this environment, reinforcement learning agents learn to trade successfully. The agents can trade the usage right of computational cores to process high-priority, high-reward jobs faster than low-priority, low-reward jobs. However, due to combinatorial effects, the action and observation spaces of a simple reinforcement learning agent in this environment scale exponentially with key parameters of the problem size. However, the exponential scaling behavior can be transformed into a linear one if the agent is split into several independent sub-units. We further improve this distributed architecture using agent-internal parameter sharing. Moreover, it can be extended to set the exchange prices autonomously. We show that in our scheduling environment, the advantages of a distributed agent architecture clearly outweigh more aggregated approaches. We demonstrate that the distributed agent architecture becomes even more performant using agent-internal parameter sharing. Finally, we investigate how two different reward functions affect autonomous pricing and the corresponding scheduling.

  Training Certifiably Robust Neural Networks Against Semantic Perturbations

• Authors: Rem Yang, Jacob Laurel, Sasa Misailovic, Gagandeep Singh
• Subjects: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
• Arxiv link: https://arxiv.org/abs/2207.11177
• Pdf link: https://arxiv.org/pdf/2207.11177
• Abstract Semantic image perturbations, such as scaling and rotation, have been shown to easily deceive deep neural networks (DNNs). Hence, training DNNs to be certifiably robust to these perturbations is critical. However, no prior work has been able to incorporate the objective of deterministic semantic robustness into the training procedure, as existing deterministic semantic verifiers are exceedingly slow. To address these challenges, we propose Certified Semantic Training (CST), the first training framework for deterministic certified robustness against semantic image perturbations. Our framework leverages a novel GPU-optimized verifier that, unlike existing works, is fast enough for use in training. Our results show that networks trained via CST consistently achieve both better provable semantic robustness and clean accuracy, compared to networks trained via baselines based on existing works.

  Keyword: calibration

  A Non-intrusive Approach for Physics-constrained Learning with Application to Fuel Cell Modeling

• Authors: Vishal Srivastava, Valentin Sulzer, Peyman Mohtat, Jason B. Siegel, Karthik Duraisamy
• Subjects: Computational Engineering, Finance, and Science (cs.CE); Data Analysis, Statistics and Probability (physics.data-an)
• Arxiv link: https://arxiv.org/abs/2207.10819
• Pdf link: https://arxiv.org/pdf/2207.10819
• Abstract A data-driven model augmentation framework, referred to as Weakly-coupled Integrated Inference and Machine Learning (IIML), is presented to improve the predictive accuracy of physical models. In contrast to parameter calibration, this work seeks corrections to the structure of the model by a) inferring augmentation fields that are consistent with the underlying model, and b) transforming these fields into corrective model forms. The proposed approach couples the inference and learning steps in a weak sense via an alternating optimization approach. This coupling ensures that the augmentation fields remain learnable and maintain consistent functional relationships with local modeled quantities across the training dataset. An iterative solution procedure is presented in this paper, removing the need to embed the augmentation function during the inference process. This framework is used to infer an augmentation introduced within a Polymer electrolyte membrane fuel cell (PEMFC) model using a small amount of training data (from only 14 training cases.) These training cases belong to a dataset consisting of high-fidelity simulation data obtained from a high-fidelity model of a first generation Toyota Mirai. All cases in this dataset are characterized by different inflow and outflow conditions on the same geometry. When tested on 1224 different configurations, the inferred augmentation significantly improves the predictive accuracy for a wide range of physical conditions. Predictions and available data for the current density distribution are also compared to demonstrate the predictive capability of the model for quantities of interest which were not involved in the inference process. The results demonstrate that the weakly-coupled IIML framework offers sophisticated and robust model augmentation capabilities without requiring extensive changes to the numerical solver.

  Bi-directional Contrastive Learning for Domain Adaptive Semantic Segmentation

• Authors: Geon Lee, Chanho Eom, Wonkyung Lee, Hyekang Park, Bumsub Ham
• Subjects: Computer Vision and Pattern Recognition (cs.CV)
• Arxiv link: https://arxiv.org/abs/2207.10892
• Pdf link: https://arxiv.org/pdf/2207.10892
• Abstract We present a novel unsupervised domain adaptation method for semantic segmentation that generalizes a model trained with source images and corresponding ground-truth labels to a target domain. A key to domain adaptive semantic segmentation is to learn domain-invariant and discriminative features without target ground-truth labels. To this end, we propose a bi-directional pixel-prototype contrastive learning framework that minimizes intra-class variations of features for the same object class, while maximizing inter-class variations for different ones, regardless of domains. Specifically, our framework aligns pixel-level features and a prototype of the same object class in target and source images (i.e., positive pairs), respectively, sets them apart for different classes (i.e., negative pairs), and performs the alignment and separation processes toward the other direction with pixel-level features in the source image and a prototype in the target image. The cross-domain matching encourages domain-invariant feature representations, while the bidirectional pixel-prototype correspondences aggregate features for the same object class, providing discriminative features. To establish training pairs for contrastive learning, we propose to generate dynamic pseudo labels of target images using a non-parametric label transfer, that is, pixel-prototype correspondences across different domains. We also present a calibration method compensating class-wise domain biases of prototypes gradually during training.

  Efficient Modeling of Future Context for Image Captioning

• Authors: Zhengcong Fei, Junshi Huang, Xiaoming Wei, Xiaolin Wei
• Subjects: Computer Vision and Pattern Recognition (cs.CV)
• Arxiv link: https://arxiv.org/abs/2207.10897
• Pdf link: https://arxiv.org/pdf/2207.10897
• Abstract Existing approaches to image captioning usually generate the sentence word-by-word from left to right, with the constraint of conditioned on local context including the given image and history generated words. There have been many studies target to make use of global information during decoding, e.g., iterative refinement. However, it is still under-explored how to effectively and efficiently incorporate the future context. To respond to this issue, inspired by that Non-Autoregressive Image Captioning (NAIC) can leverage two-side relation with modified mask operation, we aim to graft this advance to the conventional Autoregressive Image Captioning (AIC) model while maintaining the inference efficiency without extra time cost. Specifically, AIC and NAIC models are first trained combined with shared visual encoders, forcing the visual encoder to contain sufficient and valid future context; then the AIC model is encouraged to capture the causal dynamics of cross-layer interchanging from NAIC model on its unconfident words, which follows a teacher-student paradigm and optimized with the distribution calibration training objective. Empirical evidences demonstrate that our proposed approach clearly surpass the state-of-the-art baselines in both automatic metrics and human evaluations on the MS COCO benchmark. The source code is available at: https://github.com/feizc/Future-Caption.

  Open video data sharing in developmental and behavioural science

• Authors: Peter B Marschik, Tomas Kulvicius, Sarah Flügge, Claudius Widmann, Karin Nielsen-Saines, Martin Schulte-Rüther, Britta Hüning, Sven Bölte, Luise Poustka, Jeff Sigafoos, Florentin Wörgötter, Christa Einspieler, Dajie Zhang
• Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
• Arxiv link: https://arxiv.org/abs/2207.11020
• Pdf link: https://arxiv.org/pdf/2207.11020
• Abstract Video recording is a widely used method for documenting infant and child behaviours in research and clinical practice. Video data has rarely been shared due to ethical concerns of confidentiality, although the need of shared large-scaled datasets remains increasing. This demand is even more imperative when data-driven computer-based approaches are involved, such as screening tools to complement clinical assessments. To share data while abiding by privacy protection rules, a critical question arises whether efforts at data de-identification reduce data utility? We addressed this question by showcasing the Prechtl's general movements assessment (GMA), an established and globally practised video-based diagnostic tool in early infancy for detecting neurological deficits, such as cerebral palsy. To date, no shared expert-annotated large data repositories for infant movement analyses exist. Such datasets would massively benefit training and recalibration of human assessors and the development of computer-based approaches. In the current study, sequences from a prospective longitudinal infant cohort with a total of 19451 available general movements video snippets were randomly selected for human clinical reasoning and computer-based analysis. We demonstrated for the first time that pseudonymisation by face-blurring video recordings is a viable approach. The video redaction did not affect classification accuracy for either human assessors or computer vision methods, suggesting an adequate and easy-to-apply solution for sharing movement video data. We call for further explorations into efficient and privacy rule-conforming approaches for deidentifying video data in scientific and clinical fields beyond movement assessments. These approaches shall enable sharing and merging stand-alone video datasets into large data pools to advance science and public health.

  Improving Predictive Performance and Calibration by Weight Fusion in Semantic Segmentation

• Authors: Timo Sämann, Ahmed Mostafa Hammam, Andrei Bursuc, Christoph Stiller, Horst-Michael Groß
• Subjects: Computer Vision and Pattern Recognition (cs.CV)
• Arxiv link: https://arxiv.org/abs/2207.11211
• Pdf link: https://arxiv.org/pdf/2207.11211
• Abstract Averaging predictions of a deep ensemble of networks is apopular and effective method to improve predictive performance andcalibration in various benchmarks and Kaggle competitions. However, theruntime and training cost of deep ensembles grow linearly with the size ofthe ensemble, making them unsuitable for many applications. Averagingensemble weights instead of predictions circumvents this disadvantageduring inference and is typically applied to intermediate checkpoints ofa model to reduce training cost. Albeit effective, only few works haveimproved the understanding and the performance of weight averaging.Here, we revisit this approach and show that a simple weight fusion (WF)strategy can lead to a significantly improved predictive performance andcalibration. We describe what prerequisites the weights must meet interms of weight space, functional space and loss. Furthermore, we presenta new test method (called oracle test) to measure the functional spacebetween weights. We demonstrate the versatility of our WF strategy acrossstate of the art segmentation CNNs and Transformers as well as real worlddatasets such as BDD100K and Cityscapes. We compare WF with similarapproaches and show our superiority for in- and out-of-distribution datain terms of predictive performance and calibration.