PyTorchKR Papers

최신 논문

총 848편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2210.00604

Ensembling improves stability and power of feature selection for deep learning models

With the growing adoption of deep learning models in different real-world domains, including computational biology, it is often necessary to understand which data features are essential for the model's decision. Despite extensive recent efforts to define different feature importance metrics for deep learning models, we identified that inherent stochasticity in the design and training of deep learning models makes commonly used feature importance scores unstable. This results in varied explanations or selections of different features across different runs of the model. We demonstrate how the signal strength of features and correlation among features directly contribute to this instability. To address this instability, we explore the ensembling of feature importance scores of models across different epochs and find that this simple approach can substantially address this issue. For example, we consider knockoff inference as they allow feature selection with statistical guarantees. We discover considerable variability in selected features in different epochs of deep learning training, and the best selection of features doesn't necessarily occur at the lowest validation loss, the conventional approach to determine the best model. As such, we present a framework to combine the feature importance of trained models across different hyperparameter settings and epochs, and instead of selecting features from one best model, we perform an ensemble of feature importance scores from numerous good models. Across the range of experiments in simulated and various real-world datasets, we demonstrate that the proposed framework consistently improves the power of feature selection.

arXiv:2209.14881공식 코드 · Papers with Code

Sequential Attention for Feature Selection

Feature selection is the problem of selecting a subset of features for a machine learning model that maximizes model quality subject to a budget constraint. For neural networks, prior methods, including those based on $\ell_1$ regularization, attention, and other techniques, typically select the entire feature subset in one evaluation round, ignoring the residual value of features during selection, i.e., the marginal contribution of a feature given that other features have already been selected. We propose a feature selection algorithm called Sequential Attention that achieves state-of-the-art empirical results for neural networks. This algorithm is based on an efficient one-pass implementation of greedy forward selection and uses attention weights at each step as a proxy for feature importance. We give theoretical insights into our algorithm for linear regression by showing that an adaptation to this setting is equivalent to the classical Orthogonal Matching Pursuit (OMP) algorithm, and thus inherits all of its provable guarantees. Our theoretical and empirical analyses offer new explanations towards the effectiveness of attention and its connections to overparameterization, which may be of independent interest.

arXiv:2209.13858공식 코드 · Papers with Code

Variance Tolerance Factors For Interpreting ALL Neural Networks

Black box models only provide results for deep learning tasks, and lack informative details about how these results were obtained. Knowing how input variables are related to outputs, in addition to why they are related, can be critical to translating predictions into laboratory experiments, or defending a model prediction under scrutiny. In this paper, we propose a general theory that defines a variance tolerance factor (VTF) inspired by influence function, to interpret features in the context of black box neural networks by ranking the importance of features, and construct a novel architecture consisting of a base model and feature model to explore the feature importance in a Rashomon set that contains all well-performing neural networks. Two feature importance ranking methods in the Rashomon set and a feature selection method based on the VTF are created and explored. A thorough evaluation on synthetic and benchmark datasets is provided, and the method is applied to two real world examples predicting the formation of noncrystalline gold nanoparticles and the chemical toxicity 1793 aromatic compounds exposed to a protozoan ciliate for 40 hours.

arXiv:2209.10876

An Attention Matrix for Every Decision: Faithfulness-based Arbitration Among Multiple Attention-Based Interpretations of Transformers in Text Classification

Transformers are widely used in natural language processing, where they consistently achieve state-of-the-art performance. This is mainly due to their attention-based architecture, which allows them to model rich linguistic relations between (sub)words. However, transformers are difficult to interpret. Being able to provide reasoning for its decisions is an important property for a model in domains where human lives are affected. With transformers finding wide use in such fields, the need for interpretability techniques tailored to them arises. We propose a new technique that selects the most faithful attention-based interpretation among the several ones that can be obtained by combining different head, layer and matrix operations. In addition, two variations are introduced towards (i) reducing the computational complexity, thus being faster and friendlier to the environment, and (ii) enhancing the performance in multi-label data. We further propose a new faithfulness metric that is more suitable for transformer models and exhibits high correlation with the area under the precision-recall curve based on ground truth rationales. We validate the utility of our contributions with a series of quantitative and qualitative experiments on seven datasets.

arXiv:2209.11222공식 코드 · Papers with Code

Concept Activation Regions: A Generalized Framework For Concept-Based Explanations

Concept-based explanations permit to understand the predictions of a deep neural network (DNN) through the lens of concepts specified by users. Existing methods assume that the examples illustrating a concept are mapped in a fixed direction of the DNN's latent space. When this holds true, the concept can be represented by a concept activation vector (CAV) pointing in that direction. In this work, we propose to relax this assumption by allowing concept examples to be scattered across different clusters in the DNN's latent space. Each concept is then represented by a region of the DNN's latent space that includes these clusters and that we call concept activation region (CAR). To formalize this idea, we introduce an extension of the CAV formalism that is based on the kernel trick and support vector classifiers. This CAR formalism yields global concept-based explanations and local concept-based feature importance. We prove that CAR explanations built with radial kernels are invariant under latent space isometries. In this way, CAR assigns the same explanations to latent spaces that have the same geometry. We further demonstrate empirically that CARs offer (1) more accurate descriptions of how concepts are scattered in the DNN's latent space; (2) global explanations that are closer to human concept annotations and (3) concept-based feature importance that meaningfully relate concepts with each other. Finally, we use CARs to show that DNNs can autonomously rediscover known scientific concepts, such as the prostate cancer grading system.

arXiv:2209.10578

Algorithm-Agnostic Interpretations for Clustering

A clustering outcome for high-dimensional data is typically interpreted via post-processing, involving dimension reduction and subsequent visualization. This destroys the meaning of the data and obfuscates interpretations. We propose algorithm-agnostic interpretation methods to explain clustering outcomes in reduced dimensions while preserving the integrity of the data. The permutation feature importance for clustering represents a general framework based on shuffling feature values and measuring changes in cluster assignments through custom score functions. The individual conditional expectation for clustering indicates observation-wise changes in the cluster assignment due to changes in the data. The partial dependence for clustering evaluates average changes in cluster assignments for the entire feature space. All methods can be used with any clustering algorithm able to reassign instances through soft or hard labels. In contrast to common post-processing methods such as principal component analysis, the introduced methods maintain the original structure of the features.

arXiv:2209.09638

Machine Learning Models Evaluation and Feature Importance Analysis on NPL Dataset

Predicting the probability of non-performing loans for individuals has a vital and beneficial role for banks to decrease credit risk and make the right decisions before giving the loan. The trend to make these decisions are based on credit study and in accordance with generally accepted standards, loan payment history, and demographic data of the clients. In this work, we evaluate how different Machine learning models such as Random Forest, Decision tree, KNN, SVM, and XGBoost perform on the dataset provided by a private bank in Ethiopia. Further, motivated by this evaluation we explore different feature selection methods to state the important features for the bank. Our findings show that XGBoost achieves the highest F1 score on the KMeans SMOTE over-sampled data. We also found that the most important features are the age of the applicant, years of employment, and total income of the applicant rather than collateral-related features in evaluating credit risk.

arXiv:2209.08595

Low-cost machine learning approach to the prediction of transition metal phosphor excited state properties

Photoactive iridium complexes are of broad interest due to their applications ranging from lighting to photocatalysis. However, the excited state property prediction of these complexes challenges ab initio methods such as time-dependent density functional theory (TDDFT) both from an accuracy and a computational cost perspective, complicating high throughput virtual screening (HTVS). We instead leverage low-cost machine learning (ML) models to predict the excited state properties of photoactive iridium complexes. We use experimental data of 1,380 iridium complexes to train and evaluate the ML models and identify the best-performing and most transferable models to be those trained on electronic structure features from low-cost density functional theory tight binding calculations. Using these models, we predict the three excited state properties considered, mean emission energy of phosphorescence, excited state lifetime, and emission spectral integral, with accuracy competitive with or superseding TDDFT. We conduct feature importance analysis to identify which iridium complex attributes govern excited state properties and we validate these trends with explicit examples. As a demonstration of how our ML models can be used for HTVS and the acceleration of chemical discovery, we curate a set of novel hypothetical iridium complexes and identify promising ligands for the design of new phosphors.

arXiv:2209.08418

Sample-based Uncertainty Quantification with a Single Deterministic Neural Network

Development of an accurate, flexible, and numerically efficient uncertainty quantification (UQ) method is one of fundamental challenges in machine learning. Previously, a UQ method called DISCO Nets has been proposed (Bouchacourt et al., 2016), which trains a neural network by minimizing the energy score. In this method, a random noise vector in $\mathbb{R}^{10\text{--}100}$ is concatenated with the original input vector in order to produce a diverse ensemble forecast despite using a single neural network. While this method has shown promising performance on a hand pose estimation task in computer vision, it remained unexplored whether this method works as nicely for regression on tabular data, and how it competes with more recent advanced UQ methods such as NGBoost. In this paper, we propose an improved neural architecture of DISCO Nets that admits faster and more stable training while only using a compact noise vector of dimension $\sim \mathcal{O}(1)$. We benchmark this approach on miscellaneous real-world tabular datasets and confirm that it is competitive with or even superior to standard UQ baselines. Moreover we observe that it exhibits better point forecast performance than a neural network of the same size trained with the conventional mean squared error. As another advantage of the proposed method, we show that local feature importance computation methods such as SHAP can be easily applied to any subregion of the predictive distribution. A new elementary proof for the validity of using the energy score to learn predictive distributions is also provided.

arXiv:2209.04802

Examining Uniqueness and Permanence of the WAY EEG GAL dataset toward User Authentication

This study evaluates the discriminating capacity (uniqueness) of the EEG data from the WAY EEG GAL public dataset to authenticate individuals against one another as well as its permanence. In addition to the EEG data, Luciw et al. provide EMG (Electromyography), and kinematics data for engineers and researchers to utilize WAY EEG GAL for further studies. However, evaluating the EMG and kinematics data is outside the scope of this study. The goal of the state-of-the-art is to determine whether EEG data can be utilized to control prosthetic devices. On the other hand, this study aims to evaluate the separability of individuals through EEG data to perform user authentication. A feature importance algorithm is utilized to select the best features for each user to authenticate them against all others. The authentication platform implemented for this study is based on Machine Learning models/classifiers. As an initial test, two pilot studies are performed using Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM) to observe the learning trends of the models by multi-labeling the EEG dataset. Utilizing kNN first as the classifier for user authentication, accuracy around 75% is observed. Thereafter to improve the performance both linear and non-linear SVMs are used to perform classification. The overall average accuracies of 85.18% and 86.92% are achieved using linear and non-linear SVMs respectively. In addition to accuracy, F1 scores are also calculated. The overall average F1 score of 87.51% and 88.94% are achieved for linear and non-linear SVMs respectively. Beyond the overall performance, high performing individuals with 95.3% accuracy (95.3% F1 score) using linear SVM and 97.4% accuracy (97.3% F1 score) using non-linear SVM are also observed.

arXiv:2209.01607

A Case Study on the Classification of Lost Circulation Events During Drilling using Machine Learning Techniques on an Imbalanced Large Dataset

This study presents machine learning models that forecast and categorize lost circulation severity preemptively using a large class imbalanced drilling dataset. We demonstrate reproducible core techniques involved in tackling a large drilling engineering challenge utilizing easily interpretable machine learning approaches. We utilized a 65,000+ records data with class imbalance problem from Azadegan oilfield formations in Iran. Eleven of the dataset's seventeen parameters are chosen to be used in the classification of five lost circulation events. To generate classification models, we used six basic machine learning algorithms and four ensemble learning methods. Linear Discriminant Analysis (LDA), Logistic Regression (LR), Support Vector Machines (SVM), Classification and Regression Trees (CART), k-Nearest Neighbors (KNN), and Gaussian Naive Bayes (GNB) are the six fundamental techniques. We also used bagging and boosting ensemble learning techniques in the investigation of solutions for improved predicting performance. The performance of these algorithms is measured using four metrics: accuracy, precision, recall, and F1-score. The F1-score weighted to represent the data imbalance is chosen as the preferred evaluation criterion. The CART model was found to be the best in class for identifying drilling fluid circulation loss events with an average weighted F1-score of 0.9904 and standard deviation of 0.0015. Upon application of ensemble learning techniques, a Random Forest ensemble of decision trees showed the best predictive performance. It identified and classified lost circulation events with a perfect weighted F1-score of 1.0. Using Permutation Feature Importance (PFI), the measured depth was found to be the most influential factor in accurately recognizing lost circulation events while drilling.

arXiv:2209.01939

Incremental Permutation Feature Importance (iPFI): Towards Online Explanations on Data Streams

Explainable Artificial Intelligence (XAI) has mainly focused on static learning scenarios so far. We are interested in dynamic scenarios where data is sampled progressively, and learning is done in an incremental rather than a batch mode. We seek efficient incremental algorithms for computing feature importance (FI) measures, specifically, an incremental FI measure based on feature marginalization of absent features similar to permutation feature importance (PFI). We propose an efficient, model-agnostic algorithm called iPFI to estimate this measure incrementally and under dynamic modeling conditions including concept drift. We prove theoretical guarantees on the approximation quality in terms of expectation and variance. To validate our theoretical findings and the efficacy of our approaches compared to traditional batch PFI, we conduct multiple experimental studies on benchmark data with and without concept drift.

arXiv:2208.12705

A machine learning approach to predict the structural and magnetic properties of Heusler alloy families

Random forest (RF) regression model is used to predict the lattice constant, magnetic moment and formation energies of full Heusler alloys, half Heusler alloys, inverse Heusler alloys and quaternary Heusler alloys based on existing as well as indigenously prepared databases. Prior analysis was carried out to check the distribution of the data points of the response variables and found that in most of the cases, the data is not normally distributed. The outcome of the RF model performance is sufficiently accurate to predict the response variables on the test data and also shows its robustness against overfitting, outliers, multicollinearity and distribution of data points. The parity plots between the machine learning predicted values against the computed values using density functional theory (DFT) shows linear behavior with adjusted R2 values lying in the range of 0.80 to 0.94 for all the predicted properties for different types of Heusler alloys. Feature importance analysis shows that the valence electron numbers plays an important feature role in the prediction for most of the predicted outcomes. Case studies with one full Heusler alloy and one quaternary Heusler alloy were also mentioned comparing the machine learning predicted results with our earlier theoretical calculated values and experimentally measured results, suggesting high accuracy of the model predicted results.

arXiv:2208.09623

Learning to predict test effectiveness

The high cost of the test can be dramatically reduced, provided that the coverability as an inherent feature of the code under test is predictable. This article offers a machine learning model to predict the extent to which the test could cover a class in terms of a new metric called Coverageability. The prediction model consists of an ensemble of four regression models. The learning samples consist of feature vectors, where features are source code metrics computed for a class. The samples are labeled by the Coverageability values computed for their corresponding classes. We offer a mathematical model to evaluate test effectiveness in terms of size and coverage of the test suite generated automatically for each class. We extend the size of the feature space by introducing a new approach to defining sub-metrics in terms of existing source code metrics. Using feature importance analysis on the learned prediction models, we sort source code metrics in the order of their impact on the test effectiveness. As a result of which, we found the class strict cyclomatic complexity as the most influential source code metric. Our experiments with the prediction models on a large corpus of Java projects containing about 23,000 classes demonstrate the Mean Absolute Error (MAE) of 0.032, Mean Squared Error (MSE) of 0.004, and an R2-score of 0.855. Compared with the state-of-the-art coverage prediction models, our models improve MAE, MSE, and an R2-score by 5.78%, 2.84%, and 20.71%, respectively.

arXiv:2208.09614공식 코드 · Papers with Code

An ensemble meta-estimator to predict source code testability

Unlike most other software quality attributes, testability cannot be evaluated solely based on the characteristics of the source code. The effectiveness of the test suite and the budget assigned to the test highly impact the testability of the code under test. The size of a test suite determines the test effort and cost, while the coverage measure indicates the test effectiveness. Therefore, testability can be measured based on the coverage and number of test cases provided by a test suite, considering the test budget. This paper offers a new equation to estimate testability regarding the size and coverage of a given test suite. The equation has been used to label 23,000 classes belonging to 110 Java projects with their testability measure. The labeled classes were vectorized using 262 metrics. The labeled vectors were fed into a family of supervised machine learning algorithms, regression, to predict testability in terms of the source code metrics. Regression models predicted testability with an R2 of 0.68 and a mean squared error of 0.03, suitable in practice. Fifteen software metrics highly affecting testability prediction were identified using a feature importance analysis technique on the learned model. The proposed models have improved mean absolute error by 38% due to utilizing new criteria, metrics, and data compared with the relevant study on predicting branch coverage as a test criterion. As an application of testability prediction, it is demonstrated that automated refactoring of 42 smelly Java classes targeted at improving the 15 influential software metrics could elevate their testability by an average of 86.87%.

arXiv:2208.09195

Real-Time Robust Video Object Detection System Against Physical-World Adversarial Attacks

DNN-based video object detection (VOD) powers autonomous driving and video surveillance industries with rising importance and promising opportunities. However, adversarial patch attack yields huge concern in live vision tasks because of its practicality, feasibility, and powerful attack effectiveness. This work proposes Themis, a software/hardware system to defend against adversarial patches for real-time robust video object detection. We observe that adversarial patches exhibit extremely localized superficial feature importance in a small region with non-robust predictions, and thus propose the adversarial region detection algorithm for adversarial effect elimination. Themis also proposes a systematic design to efficiently support the algorithm by eliminating redundant computations and memory traffics. Experimental results show that the proposed methodology can effectively recover the system from the adversarial attack with negligible hardware overhead.

arXiv:2208.08858

Estimating Sunlight Using GNSS Signal Strength from Smartphone

Excessive or inadequate exposure to ultraviolet light (UV) is harmful to health and causes osteoporosis, colon cancer, and skin cancer. The UV Index, a standard scale of UV light, tends to increase in sunny places and sharply decrease in the shade. A method for distinguishing shady and sunny places would help us to prevent and cure diseases caused by UV. However, the existing methods, such as carrying UV sensors, impose a load on the user, whereas city-level UV forecasts do not have enough granularity for monitoring an individual's UV exposure. This paper proposes a method to detect sunny and shady places by using an off-the-shelf mobile device. The method detects these places by using a characteristic of the GNSS signal strength that is attenuated by objects around the device. As a dataset, we collected GNSS signal data, such as C/N0, satellite ID, satellite angle, and sun angle, together with reference data (i.e., sunny and shady place information every minute) for four days from five locations. Using the dataset, we created twelve classification models by using supervised machine learning methods and evaluated their performance by 4-fold cross-validation. In addition, we investigated the feature importance and the effect of combining features. The performance evaluation showed that our classification model could classify sunny and shady places with more than 97% accuracy in the best case. Moreover, our investigation revealed that the value of C/N0 at a moment and its time series (i.e., C/N0 value before and after the moment) are more important features.

arXiv:2208.06151공식 코드 · Papers with Code

Unifying local and global model explanations by functional decomposition of low dimensional structures

We consider a global representation of a regression or classification function by decomposing it into the sum of main and interaction components of arbitrary order. We propose a new identification constraint that allows for the extraction of interventional SHAP values and partial dependence plots, thereby unifying local and global explanations. With our proposed identification, a feature's partial dependence plot corresponds to the main effect term plus the intercept. The interventional SHAP value of feature $k$ is a weighted sum of the main component and all interaction components that include $k$, with the weights given by the reciprocal of the component's dimension. This brings a new perspective to local explanations such as SHAP values which were previously motivated by game theory only. We show that the decomposition can be used to reduce direct and indirect bias by removing all components that include a protected feature. Lastly, we motivate a new measure of feature importance. In principle, our proposed functional decomposition can be applied to any machine learning model, but exact calculation is only feasible for low-dimensional structures or ensembles of those. We provide an algorithm and efficient implementation for gradient-boosted trees (xgboost) and random planted forest. Conducted experiments suggest that our method provides meaningful explanations and reveals interactions of higher orders. The proposed methods are implemented in an R package, available at \url{https://github.com/PlantedML/glex}.

arXiv:2208.04732공식 코드 · Papers with Code

Efficient Novelty Detection Methods for Early Warning of Potential Fatal Diseases

Fatal diseases, as Critical Health Episodes (CHEs), represent real dangers for patients hospitalized in Intensive Care Units. These episodes can lead to irreversible organ damage and death. Nevertheless, diagnosing them in time would greatly reduce their inconvenience. This study therefore focused on building a highly effective early warning system for CHEs such as Acute Hypotensive Episodes and Tachycardia Episodes. To facilitate the precocity of the prediction, a gap of one hour was considered between the observation periods (Observation Windows) and the periods during which a critical event can occur (Target Windows). The MIMIC II dataset was used to evaluate the performance of the proposed system. This system first includes extracting additional features using three different modes. Then, the feature selection process allowing the selection of the most relevant features was performed using the Mutual Information Gain feature importance. Finally, the high-performance predictive model LightGBM was used to perform episode classification. This approach called MIG-LightGBM was evaluated using five different metrics: Event Recall (ER), Reduced Precision (RP), average Anticipation Time (aveAT), average False Alarms (aveFA), and Event F1-score (EF1-score). A method is therefore considered highly efficient for the early prediction of CHEs if it exhibits not only a large aveAT but also a large EF1-score and a low aveFA. Compared to systems using Extreme Gradient Boosting, Support Vector Classification or Naive Bayes as a predictive model, the proposed system was found to be highly dominant. It also confirmed its superiority over the Layered Learning approach.

arXiv:2208.04343공식 코드 · Papers with Code

EFI: A Toolbox for Feature Importance Fusion and Interpretation in Python

This paper presents an open-source Python toolbox called Ensemble Feature Importance (EFI) to provide machine learning (ML) researchers, domain experts, and decision makers with robust and accurate feature importance quantification and more reliable mechanistic interpretation of feature importance for prediction problems using fuzzy sets. The toolkit was developed to address uncertainties in feature importance quantification and lack of trustworthy feature importance interpretation due to the diverse availability of machine learning algorithms, feature importance calculation methods, and dataset dependencies. EFI merges results from multiple machine learning models with different feature importance calculation approaches using data bootstrapping and decision fusion techniques, such as mean, majority voting and fuzzy logic. The main attributes of the EFI toolbox are: (i) automatic optimisation of ML algorithms, (ii) automatic computation of a set of feature importance coefficients from optimised ML algorithms and feature importance calculation techniques, (iii) automatic aggregation of importance coefficients using multiple decision fusion techniques, and (iv) fuzzy membership functions that show the importance of each feature to the prediction task. The key modules and functions of the toolbox are described, and a simple example of their application is presented using the popular Iris dataset.