Abstract
Learning-based control schemes have recently shown great efficacy performing complex tasks. However, in order to deploy them in real systems, it is of vital importance to guarantee that the system will remain safe during online training and execution. We therefore need safe online learning frameworks able to autonomously reason about whether the current information at their disposal is enough to ensure safety or, in contrast, new measurements are required. In this paper, we present a framework consisting of two parts: first, an out-of-distribution detection mechanism actively collecting measurements when needed to guarantee that at least one safety backup direction is always available for use; and second, a Gaussian Process-based probabilistic safety-critical controller that ensures the system stays safe at all times with high probability. Our method exploits model knowledge through the use of Control Barrier Functions, and collects measurements from the stream of online data in an event-triggered fashion to guarantee recursive feasibility of the learned safety-critical controller. This, in turn, allows us to provide formal results of forward invariance of a safe set with high probability, even in a priori unexplored regions. Finally, we validate the proposed framework in numerical simulations of an adaptive cruise control system.
Keyword: expected calibration error
There is no result
Keyword: overconfident
"Am I Private and If So, how Many?" - Communicating Privacy Guarantees of Differential Privacy with Risk Communication Formats
Authors: Daniel Franzen (1), Saskia Nuñez von Voigt (2), Peter Sörries (1), Florian Tschorsch (2), Claudia Müller-Birn (1) ((1) Freie Universität Berlin, (2) Technische Universität Berlin)
Subjects: Human-Computer Interaction (cs.HC); Cryptography and Security (cs.CR)
Abstract
Decisions about sharing personal information are not trivial, since there are many legitimate and important purposes for such data collection, but often the collected data can reveal sensitive information about individuals. Privacy-preserving technologies, such as differential privacy (DP), can be employed to protect the privacy of individuals and, furthermore, provide mathematically sound guarantees on the maximum privacy risk. However, they can only support informed privacy decisions, if individuals understand the provided privacy guarantees. This article proposes a novel approach for communicating privacy guarantees to support individuals in their privacy decisions when sharing data. For this, we adopt risk communication formats from the medical domain in conjunction with a model for privacy guarantees of DP to create quantitative privacy risk notifications. We conducted a crowd-sourced study with 343 participants to evaluate how well our notifications conveyed the privacy risk information and how confident participants were about their own understanding of the privacy risk. Our findings suggest that these new notifications can communicate the objective information similarly well to currently used qualitative notifications, but left individuals less confident in their understanding. We also discovered that several of our notifications and the currently used qualitative notification disadvantage individuals with low numeracy: these individuals appear overconfident compared to their actual understanding of the associated privacy risks and are, therefore, less likely to seek the needed additional information before an informed decision. The promising results allow for multiple directions in future research, for example, adding visual aids or tailoring privacy risk communication to characteristics of the individuals.
Keyword: overconfidence
There is no result
Keyword: confidence
Some Supervision Required: Incorporating Oracle Policies in Reinforcement Learning via Epistemic Uncertainty Metrics
Authors: Jun Jet Tai, Jordan K. Terry, Mauro S. Innocente, James Brusey, Nadjim Horri
Abstract
An inherent problem in reinforcement learning is coping with policies that are uncertain about what action to take (or the value of a state). Model uncertainty, more formally known as epistemic uncertainty, refers to the expected prediction error of a model beyond the sampling noise. In this paper, we propose a metric for epistemic uncertainty estimation in Q-value functions, which we term pathwise epistemic uncertainty. We further develop a method to compute its approximate upper bound, which we call F -value. We experimentally apply the latter to Deep Q-Networks (DQN) and show that uncertainty estimation in reinforcement learning serves as a useful indication of learning progress. We then propose a new approach to improving sample efficiency in actor-critic algorithms by learning from an existing (previously learned or hard-coded) oracle policy while uncertainty is high, aiming to avoid unproductive random actions during training. We term this Critic Confidence Guided Exploration (CCGE). We implement CCGE on Soft Actor-Critic (SAC) using our F-value metric, which we apply to a handful of popular Gym environments and show that it achieves better sample efficiency and total episodic reward than vanilla SAC in limited contexts.
Individual Tree Detection in Large-Scale Urban Environments using High-Resolution Multispectral Imagery
Authors: Jonathan Ventura, Milo Honsberger, Cameron Gonsalves, Julian Rice, Camille Pawlak, Natalie L.R. Love, Skyler Han, Viet Nguyen, Keilana Sugano, Jacqueline Doremus, G. Andrew Fricker, Jenn Yost, Matt Ritter
Subjects: Computer Vision and Pattern Recognition (cs.CV)
Abstract
We introduce a novel deep learning method for detection of individual trees in urban environments using high-resolution multispectral aerial imagery. We use a convolutional neural network to regress a confidence map indicating the locations of individual trees, which are localized using a peak finding algorithm. Our method provides complete spatial coverage by detecting trees in both public and private spaces, and can scale to very large areas. In our study area spanning five cities in Southern California, we achieved an F-score of 0.735 and an RMSE of 2.157 m. We used our method to produce a map of all trees in the urban forest of California, indicating the potential for our method to support future urban forestry studies at unprecedented scales.
Efficient Self-Supervision using Patch-based Contrastive Learning for Histopathology Image Segmentation
Authors: Nicklas Boserup, Raghavendra Selvan
Subjects: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
Abstract
Learning discriminative representations of unlabelled data is a challenging task. Contrastive self-supervised learning provides a framework to learn meaningful representations using learned notions of similarity measures from simple pretext tasks. In this work, we propose a simple and efficient framework for self-supervised image segmentation using contrastive learning on image patches, without using explicit pretext tasks or any further labeled fine-tuning. A fully convolutional neural network (FCNN) is trained in a self-supervised manner to discern features in the input images and obtain confidence maps which capture the network's belief about the objects belonging to the same class. Positive- and negative- patches are sampled based on the average entropy in the confidence maps for contrastive learning. Convergence is assumed when the information separation between the positive patches is small, and the positive-negative pairs is large. We evaluate this method for the task of segmenting nuclei from multiple histopathology datasets, and show comparable performance with relevant self-supervised and supervised methods. The proposed model only consists of a simple FCNN with 10.8k parameters and requires about 5 minutes to converge on the high resolution microscopy datasets, which is orders of magnitude smaller than the relevant self-supervised methods to attain similar performance.
A Comprehensive Study of Real-Time Object Detection Networks Across Multiple Domains: A Survey
Abstract
Deep neural network based object detectors are continuously evolving and are used in a multitude of applications, each having its own set of requirements. While safety-critical applications need high accuracy and reliability, low-latency tasks need resource and energy-efficient networks. Real-time detectors, which are a necessity in high-impact real-world applications, are continuously proposed, but they overemphasize the improvements in accuracy and speed while other capabilities such as versatility, robustness, resource and energy efficiency are omitted. A reference benchmark for existing networks does not exist, nor does a standard evaluation guideline for designing new networks, which results in ambiguous and inconsistent comparisons. We, thus, conduct a comprehensive study on multiple real-time detectors (anchor-, keypoint-, and transformer-based) on a wide range of datasets and report results on an extensive set of metrics. We also study the impact of variables such as image size, anchor dimensions, confidence thresholds, and architecture layers on the overall performance. We analyze the robustness of detection networks against distribution shifts, natural corruptions, and adversarial attacks. Also, we provide a calibration analysis to gauge the reliability of the predictions. Finally, to highlight the real-world impact, we conduct two unique case studies, on autonomous driving and healthcare applications. To further gauge the capability of networks in critical real-time applications, we report the performance after deploying the detection networks on edge devices. Our extensive empirical study can act as a guideline for the industrial community to make an informed choice on the existing networks. We also hope to inspire the research community towards a new direction in the design and evaluation of networks that focuses on a bigger and holistic overview for a far-reaching impact.
Research Note on Uncertain Probabilities and Abstract Argumentation
Authors: Pietro Baroni, Federico Cerutti, Massimiliano Giacomin, Lance M. Kaplan, Murat Sensoy
Abstract
The sixth assessment of the international panel on climate change (IPCC) states that "cumulative net CO2 emissions over the last decade (2010-2019) are about the same size as the 11 remaining carbon budget likely to limit warming to 1.5C (medium confidence)." Such reports directly feed the public discourse, but nuances such as the degree of belief and of confidence are often lost. In this paper, we propose a formal account for allowing such degrees of belief and the associated confidence to be used to label arguments in abstract argumentation settings. Differently from other proposals in probabilistic argumentation, we focus on the task of probabilistic inference over a chosen query building upon Sato's distribution semantics which has been already shown to encompass a variety of cases including the semantics of Bayesian networks. Borrowing from the vast literature on such semantics, we examine how such tasks can be dealt with in practice when considering uncertain probabilities, and discuss the connections with existing proposals for probabilistic argumentation.
Evaluating Out-of-Distribution Detectors Through Adversarial Generation of Outliers
Authors: Sangwoong Yoon, Jinwon Choi, Yonghyeon Lee, Yung-Kyun Noh, Frank Chongwoo Park
Subjects: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
Abstract
A reliable evaluation method is essential for building a robust out-of-distribution (OOD) detector. Current robustness evaluation protocols for OOD detectors rely on injecting perturbations to outlier data. However, the perturbations are unlikely to occur naturally or not relevant to the content of data, providing a limited assessment of robustness. In this paper, we propose Evaluation-via-Generation for OOD detectors (EvG), a new protocol for investigating the robustness of OOD detectors under more realistic modes of variation in outliers. EvG utilizes a generative model to synthesize plausible outliers, and employs MCMC sampling to find outliers misclassified as in-distribution with the highest confidence by a detector. We perform a comprehensive benchmark comparison of the performance of state-of-the-art OOD detectors using EvG, uncovering previously overlooked weaknesses.
Evaluate Confidence Instead of Perplexity for Zero-shot Commonsense Reasoning
Abstract
Commonsense reasoning is an appealing topic in natural language processing (NLP) as it plays a fundamental role in supporting the human-like actions of NLP systems. With large-scale language models as the backbone, unsupervised pre-training on numerous corpora shows the potential to capture commonsense knowledge. Current pre-trained language model (PLM)-based reasoning follows the traditional practice using perplexity metric. However, commonsense reasoning is more than existing probability evaluation, which is biased by word frequency. This paper reconsiders the nature of commonsense reasoning and proposes a novel commonsense reasoning metric, Non-Replacement Confidence (NRC). In detail, it works on PLMs according to the Replaced Token Detection (RTD) pre-training objective in ELECTRA, in which the corruption detection objective reflects the confidence on contextual integrity that is more relevant to commonsense reasoning than existing probability. Our proposed novel method boosts zero-shot performance on two commonsense reasoning benchmark datasets and further seven commonsense question-answering datasets. Our analysis shows that pre-endowed commonsense knowledge, especially for RTD-based PLMs, is essential in downstream reasoning.
Abstract
Proof of Work (PoW) blockchains are susceptible to adversarial majority mining attacks in the early stages due to incipient participation and corresponding low net hash power. Bootstrapping ensures safety and liveness during the transient stage by protecting against a majority mining attack, allowing a PoW chain to grow the participation base and corresponding mining hash power. Liveness is especially important since a loss of liveness will lead to loss of honest mining rewards, decreasing honest participation, hence creating an undesired spiral; indeed existing bootstrapping mechanisms offer especially weak liveness guarantees. In this paper, we propose Advocate, a new bootstrapping methodology, which achieves two main results: (a) optimal liveness and low latency under a super-majority adversary for the Nakamoto longest chain protocol and (b) immediate black-box generalization to a variety of parallel-chain based scaling architectures, including OHIE and Prism. We demonstrate via a full-stack implementation the robustness of Advocate under a 90% adversarial majority.
Scalable Hybrid Classification-Regression Solution for High-Frequency Nonintrusive Load Monitoring
Authors: Govind Saraswat, Blake Lundstrom, Murti V Salapaka
Subjects: Machine Learning (cs.LG); Systems and Control (eess.SY)
Abstract
Residential buildings with the ability to monitor and control their net-load (sum of load and generation) can provide valuable flexibility to power grid operators. We present a novel multiclass nonintrusive load monitoring (NILM) approach that enables effective net-load monitoring capabilities at high-frequency with minimal additional equipment and cost. The proposed machine learning based solution provides accurate multiclass state predictions while operating at a faster timescale (able to provide a prediction for each 60-Hz ac cycle used in US power grid) without relying on event-detection techniques. We also introduce an innovative hybrid classification-regression method that allows for the prediction of not only load on/off states via classification but also individual load operating power levels via regression. A test bed with eight residential appliances is used for validating the NILM approach. Results show that the overall method has high accuracy and, good scaling and generalization properties. Furthermore, the method is shown to have sufficient response time (within 160ms, corresponding to 10 ac cycles) to support building grid-interactive control at fast timescales relevant to the provision of grid frequency support services.
Keyword: calibration
Improving Speech Emotion Recognition Through Focus and Calibration Attention Mechanisms
Abstract
Attention has become one of the most commonly used mechanisms in deep learning approaches. The attention mechanism can help the system focus more on the feature space's critical regions. For example, high amplitude regions can play an important role for Speech Emotion Recognition (SER). In this paper, we identify misalignments between the attention and the signal amplitude in the existing multi-head self-attention. To improve the attention area, we propose to use a Focus-Attention (FA) mechanism and a novel Calibration-Attention (CA) mechanism in combination with the multi-head self-attention. Through the FA mechanism, the network can detect the largest amplitude part in the segment. By employing the CA mechanism, the network can modulate the information flow by assigning different weights to each attention head and improve the utilization of surrounding contexts. To evaluate the proposed method, experiments are performed with the IEMOCAP and RAVDESS datasets. Experimental results show that the proposed framework significantly outperforms the state-of-the-art approaches on both datasets.
Automatic Calibration of a Six-Degrees-of-Freedom Pose Estimation System
Authors: Wouter Jansen, Dennis Laurijssen, Walter Daems, Jan Steckel
Subjects: Robotics (cs.RO); Signal Processing (eess.SP)
Abstract
Systems for estimating the six-degrees-of-freedom human body pose have been improving for over two decades. Technologies such as motion capture cameras, advanced gaming peripherals and more recently both deep learning techniques and virtual reality systems have shown impressive results. However, most systems that provide high accuracy and high precision are expensive and not easy to operate. Recently, research has been carried out to estimate the human body pose using the HTC Vive virtual reality system. This system shows accurate results while keeping the cost under a 1000 USD. This system uses an optical approach. Two transmitter devices emit infrared pulses and laser planes are tracked by use of photo diodes on receiver hardware. A system using these transmitter devices combined with low-cost custom-made receiver hardware was developed previously but requires manual measurement of the position and orientation of the transmitter devices. These manual measurements can be time consuming, prone to error and not possible in particular setups. We propose an algorithm to automatically calibrate the poses of the transmitter devices in any chosen environment with custom receiver/calibration hardware. Results show that the calibration works in a variety of setups while being more accurate than what manual measurements would allow. Furthermore, the calibration movement and speed has no noticeable influence on the precision of the results.
A Comprehensive Study of Real-Time Object Detection Networks Across Multiple Domains: A Survey
Abstract
Deep neural network based object detectors are continuously evolving and are used in a multitude of applications, each having its own set of requirements. While safety-critical applications need high accuracy and reliability, low-latency tasks need resource and energy-efficient networks. Real-time detectors, which are a necessity in high-impact real-world applications, are continuously proposed, but they overemphasize the improvements in accuracy and speed while other capabilities such as versatility, robustness, resource and energy efficiency are omitted. A reference benchmark for existing networks does not exist, nor does a standard evaluation guideline for designing new networks, which results in ambiguous and inconsistent comparisons. We, thus, conduct a comprehensive study on multiple real-time detectors (anchor-, keypoint-, and transformer-based) on a wide range of datasets and report results on an extensive set of metrics. We also study the impact of variables such as image size, anchor dimensions, confidence thresholds, and architecture layers on the overall performance. We analyze the robustness of detection networks against distribution shifts, natural corruptions, and adversarial attacks. Also, we provide a calibration analysis to gauge the reliability of the predictions. Finally, to highlight the real-world impact, we conduct two unique case studies, on autonomous driving and healthcare applications. To further gauge the capability of networks in critical real-time applications, we report the performance after deploying the detection networks on edge devices. Our extensive empirical study can act as a guideline for the industrial community to make an informed choice on the existing networks. We also hope to inspire the research community towards a new direction in the design and evaluation of networks that focuses on a bigger and holistic overview for a far-reaching impact.
Keyword: out of distribution detection
There is no result
Keyword: out-of-distribution detection
Probabilistic Safe Online Learning with Control Barrier Functions
Keyword: expected calibration error
There is no result
Keyword: overconfident
"Am I Private and If So, how Many?" - Communicating Privacy Guarantees of Differential Privacy with Risk Communication Formats
Keyword: overconfidence
There is no result
Keyword: confidence
Some Supervision Required: Incorporating Oracle Policies in Reinforcement Learning via Epistemic Uncertainty Metrics
Individual Tree Detection in Large-Scale Urban Environments using High-Resolution Multispectral Imagery
Efficient Self-Supervision using Patch-based Contrastive Learning for Histopathology Image Segmentation
A Comprehensive Study of Real-Time Object Detection Networks Across Multiple Domains: A Survey
Research Note on Uncertain Probabilities and Abstract Argumentation
Evaluating Out-of-Distribution Detectors Through Adversarial Generation of Outliers
Evaluate Confidence Instead of Perplexity for Zero-shot Commonsense Reasoning
Keyword: scaling
Optimal Bootstrapping of PoW Blockchains
Scalable Hybrid Classification-Regression Solution for High-Frequency Nonintrusive Load Monitoring
Keyword: calibration
Improving Speech Emotion Recognition Through Focus and Calibration Attention Mechanisms
Automatic Calibration of a Six-Degrees-of-Freedom Pose Estimation System
A Comprehensive Study of Real-Time Object Detection Networks Across Multiple Domains: A Survey