Purpose: Despite the potential of machine learning models, the lack of generalizability has hindered their widespread adoption in clinical practice. We investigate three methodological pitfalls: (1) violation of independence assumption, (2) model evaluation with an inappropriate performance indicator or baseline for comparison, and (3) batch effect. Materials and Methods: Using several retrospective datasets, we implement machine learning models with and without the pitfalls to quantitatively illustrate these pitfalls' effect on model generalizability. Results: Violation of independence assumption, more specifically, applying oversampling, feature selection, and data augmentation before splitting data into train, validation, and test sets, respectively, led to misleading and superficial gains in F1 scores of 71.2% in predicting local recurrence and 5.0% in predicting 3-year overall survival in head and neck cancer as well as 46.0% in distinguishing histopathological patterns in lung cancer. Further, randomly distributing data points for a subject across training, validation, and test sets led to a 21.8% superficial increase in F1 score. Also, we showed the importance of the choice of performance measures and baseline for comparison. In the presence of batch effect, a model built for pneumonia detection led to F1 score of 98.7%. However, when the same model was applied to a new dataset of normal patients, it only correctly classified 3.86% of the samples. Conclusions: These methodological pitfalls cannot be captured using internal model evaluation, and the inaccurate predictions made by such models may lead to wrong conclusions and interpretations. Therefore, understanding and avoiding these pitfalls is necessary for developing generalizable models.
相關內容
In recent years, Graph Neural Networks have reported outstanding performance in tasks like community detection, molecule classification and link prediction. However, the black-box nature of these models prevents their application in domains like health and finance, where understanding the models' decisions is essential. Counterfactual Explanations (CE) provide these understandings through examples. Moreover, the literature on CE is flourishing with novel explanation methods which are tailored to graph learning. In this survey, we analyse the existing Graph Counterfactual Explanation methods, by providing the reader with an organisation of the literature according to a uniform formal notation for definitions, datasets, and metrics, thus, simplifying potential comparisons w.r.t to the method advantages and disadvantages. We discussed seven methods and sixteen synthetic and real datasets providing details on the possible generation strategies. We highlight the most common evaluation strategies and formalise nine of the metrics used in the literature. We first introduce the evaluation framework GRETEL and how it is possible to extend and use it while providing a further dimension of comparison encompassing reproducibility aspects. Finally, we provide a discussion on how counterfactual explanation interplays with privacy and fairness, before delving into open challenges and future works.
Capsule networks (see e.g. Hinton et al., 2018) aim to encode knowledge of and reason about the relationship between an object and its parts. In this paper we specify a generative model for such data, and derive a variational algorithm for inferring the transformation of each model object in a scene, and the assignments of observed parts to the objects. We derive a learning algorithm for the object models, based on variational expectation maximization (Jordan et al., 1999). We also study an alternative inference algorithm based on the RANSAC method of Fischler and Bolles (1981). We apply these inference methods to (i) data generated from multiple geometric objects like squares and triangles ("constellations"), and (ii) data from a parts-based model of faces. Recent work by Kosiorek et al. (2019) has used amortized inference via stacked capsule autoencoders (SCAEs) to tackle this problem -- our results show that we significantly outperform them where we can make comparisons (on the constellations data).
Today, computer systems hold large amounts of personal data. Yet while such an abundance of data allows breakthroughs in artificial intelligence, and especially machine learning (ML), its existence can be a threat to user privacy, and it can weaken the bonds of trust between humans and AI. Recent regulations now require that, on request, private information about a user must be removed from both computer systems and from ML models, i.e. ``the right to be forgotten''). While removing data from back-end databases should be straightforward, it is not sufficient in the AI context as ML models often `remember' the old data. Contemporary adversarial attacks on trained models have proven that we can learn whether an instance or an attribute belonged to the training data. This phenomenon calls for a new paradigm, namely machine unlearning, to make ML models forget about particular data. It turns out that recent works on machine unlearning have not been able to completely solve the problem due to the lack of common frameworks and resources. Therefore, this paper aspires to present a comprehensive examination of machine unlearning's concepts, scenarios, methods, and applications. Specifically, as a category collection of cutting-edge studies, the intention behind this article is to serve as a comprehensive resource for researchers and practitioners seeking an introduction to machine unlearning and its formulations, design criteria, removal requests, algorithms, and applications. In addition, we aim to highlight the key findings, current trends, and new research areas that have not yet featured the use of machine unlearning but could benefit greatly from it. We hope this survey serves as a valuable resource for ML researchers and those seeking to innovate privacy technologies. Our resources are publicly available at //github.com/tamlhp/awesome-machine-unlearning.
The "Right to be Forgotten" rule in machine learning (ML) practice enables some individual data to be deleted from a trained model, as pursued by recently developed machine unlearning techniques. To truly comply with the rule, a natural and necessary step is to verify if the individual data are indeed deleted after unlearning. Yet, previous parameter-space verification metrics may be easily evaded by a distrustful model trainer. Thus, Thudi et al. recently present a call to action on algorithm-level verification in USENIX Security'22. We respond to the call, by reconsidering the unlearning problem in the scenario of machine learning as a service (MLaaS), and proposing a new definition framework for Proof of Unlearning (PoUL) on algorithm level. Specifically, our PoUL definitions (i) enforce correctness properties on both the pre and post phases of unlearning, so as to prevent the state-of-the-art forging attacks; (ii) highlight proper practicality requirements of both the prover and verifier sides with minimal invasiveness to the off-the-shelf service pipeline and computational workloads. Under the definition framework, we subsequently present a trusted hardware-empowered instantiation using SGX enclave, by logically incorporating an authentication layer for tracing the data lineage with a proving layer for supporting the audit of learning. We customize authenticated data structures to support large out-of-enclave storage with simple operation logic, and meanwhile, enable proving complex unlearning logic with affordable memory footprints in the enclave. We finally validate the feasibility of the proposed instantiation with a proof-of-concept implementation and multi-dimensional performance evaluation.
Current Natural Language Inference (NLI) models achieve impressive results, sometimes outperforming humans when evaluating on in-distribution test sets. However, as these models are known to learn from annotation artefacts and dataset biases, it is unclear to what extent the models are learning the task of NLI instead of learning from shallow heuristics in their training data. We address this issue by introducing a logical reasoning framework for NLI, creating highly transparent model decisions that are based on logical rules. Unlike prior work, we show that improved interpretability can be achieved without decreasing the predictive accuracy. We almost fully retain performance on SNLI, while also identifying the exact hypothesis spans that are responsible for each model prediction. Using the e-SNLI human explanations, we verify that our model makes sensible decisions at a span level, despite not using any span labels during training. We can further improve model performance and span-level decisions by using the e-SNLI explanations during training. Finally, our model is more robust in a reduced data setting. When training with only 1,000 examples, out-of-distribution performance improves on the MNLI matched and mismatched validation sets by 13% and 16% relative to the baseline. Training with fewer observations yields further improvements, both in-distribution and out-of-distribution.
De-noising plays a crucial role in the post-processing of spectra. Machine learning-based methods show good performance in extracting intrinsic information from noisy data, but often require a high-quality training set that is typically inaccessible in real experimental measurements. Here, using spectra in angle-resolved photoemission spectroscopy (ARPES) as an example, we develop a de-noising method for extracting intrinsic spectral information without the need for a training set. This is possible as our method leverages the self-correlation information of the spectra themselves. It preserves the intrinsic energy band features and thus facilitates further analysis and processing. Moreover, since our method is not limited by specific properties of the training set compared to previous ones, it may well be extended to other fields and application scenarios where obtaining high-quality multidimensional training data is challenging.
Next generation technologies such as smart healthcare, self-driving cars, and smart cities require new approaches to deal with the network traffic generated by the Internet of Things (IoT) devices, as well as efficient programming models to deploy machine learning techniques. Serverless edge computing is an emerging computing paradigm from the integration of two recent technologies, edge computing and serverless computing, that can possibly address these challenges. However, there is little work to explore the capability and performance of such a technology. In this paper, a comprehensive performance analysis of a serverless edge computing system using popular open-source frameworks, namely, Kubeless, OpenFaaS, Fission, and funcX is presented. The experiments considered different programming languages, workloads, and the number of concurrent users. The machine learning workloads have been used to evaluate the performance of the system under different working conditions to provide insights into the best practices. The evaluation results revealed some of the current challenges in serverless edge computing and open research opportunities in this emerging technology for machine learning applications.
With advancements in computer vision taking place day by day, recently a lot of light is being shed on activity recognition. With the range for real-world applications utilizing this field of study increasing across a multitude of industries such as security and healthcare, it becomes crucial for businesses to distinguish which machine learning methods perform better than others in the area. This paper strives to aid in this predicament i.e. building upon previous related work, it employs both classical and ensemble approaches on rich pose estimation (OpenPose) and HAR datasets. Making use of appropriate metrics to evaluate the performance for each model, the results show that overall, random forest yields the highest accuracy in classifying ADLs. Relatively all the models have excellent performance across both datasets, except for logistic regression and AdaBoost perform poorly in the HAR one. With the limitations of this paper also discussed in the end, the scope for further research is vast, which can use this paper as a base in aims of producing better results.
Small data challenges have emerged in many learning problems, since the success of deep neural networks often relies on the availability of a huge amount of labeled data that is expensive to collect. To address it, many efforts have been made on training complex models with small data in an unsupervised and semi-supervised fashion. In this paper, we will review the recent progresses on these two major categories of methods. A wide spectrum of small data models will be categorized in a big picture, where we will show how they interplay with each other to motivate explorations of new ideas. We will review the criteria of learning the transformation equivariant, disentangled, self-supervised and semi-supervised representations, which underpin the foundations of recent developments. Many instantiations of unsupervised and semi-supervised generative models have been developed on the basis of these criteria, greatly expanding the territory of existing autoencoders, generative adversarial nets (GANs) and other deep networks by exploring the distribution of unlabeled data for more powerful representations. While we focus on the unsupervised and semi-supervised methods, we will also provide a broader review of other emerging topics, from unsupervised and semi-supervised domain adaptation to the fundamental roles of transformation equivariance and invariance in training a wide spectrum of deep networks. It is impossible for us to write an exclusive encyclopedia to include all related works. Instead, we aim at exploring the main ideas, principles and methods in this area to reveal where we are heading on the journey towards addressing the small data challenges in this big data era.
Machine-learning models have demonstrated great success in learning complex patterns that enable them to make predictions about unobserved data. In addition to using models for prediction, the ability to interpret what a model has learned is receiving an increasing amount of attention. However, this increased focus has led to considerable confusion about the notion of interpretability. In particular, it is unclear how the wide array of proposed interpretation methods are related, and what common concepts can be used to evaluate them. We aim to address these concerns by defining interpretability in the context of machine learning and introducing the Predictive, Descriptive, Relevant (PDR) framework for discussing interpretations. The PDR framework provides three overarching desiderata for evaluation: predictive accuracy, descriptive accuracy and relevancy, with relevancy judged relative to a human audience. Moreover, to help manage the deluge of interpretation methods, we introduce a categorization of existing techniques into model-based and post-hoc categories, with sub-groups including sparsity, modularity and simulatability. To demonstrate how practitioners can use the PDR framework to evaluate and understand interpretations, we provide numerous real-world examples. These examples highlight the often under-appreciated role played by human audiences in discussions of interpretability. Finally, based on our framework, we discuss limitations of existing methods and directions for future work. We hope that this work will provide a common vocabulary that will make it easier for both practitioners and researchers to discuss and choose from the full range of interpretation methods.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191