Programming-by-example for data transformation to improve machine learning performance

Data transformation is a laborious and time-consuming task for analysts. Programming by example (PBE) is a technique that can simplify this difficult task for data analysts by automatically generating programs for data transformation. Most of the previously proposed PBE methods are based on search algorithms, but recent improvements in machine learning (ML) have led to its application in PBE research. For example, RobustFill was proposed as an ML-based PBE method for string transformation by using long short-term memory (LSTM) as the sequential encoder–decoder model. However, an ML-based PBE method has not been developed for tabular transformations, which are used frequently in data analysis. Thus, in the present study, we propose an ML-based PBE method for tabular transformations. First, we consider the features of tabular transformations, which are more complex and data intensive than string transformations, and propose a new ML-based PBE method using the state-of-the-art Transformer sequential encoder–decoder model. To our knowledge, this is the first ML-based PBE method for tabular transformations. We also propose two decoding methods comprising multistep beam search and program validation-beam search, which are optimized for program generation, and thus generate correct programs with higher accuracy. Our evaluation results demonstrated that the Transformer-based PBE model performed much better than LSTM-based PBE when applied to tabular transformations. Furthermore, the Transformer-based model with the proposed decoding method performed better than the conventional PBE model using the search-based method.

Errors in data are a key challenge in modern data management and processing systems. Monitoring and mitigating risks associated with errors in data transformations and downstream applications, such as Machine Learning (ML) model training, requires a profound understanding of error generation and impact of errors on data pipelines. Unfortunately, scientific progress in the field is facing two main challenges: For one, research on data errors often does not adhere to the FAIR (Findable, Accessible, Interoperable, and Reusable) principles, which impedes reproducibility and comparisons. Second, existing data error models are oversimplified and fail to capture the complex statistical dependencies underlying the types and distributions of errors observed in real-world data. Building on prior work in the database management systems and statistics literature, we extend the theory on missing values to encompass a broader range of errors in tables and provide an overview of relevant error types. Combining error sampling mechanisms often observed in real data with a comprehensive categorization of errors, we introduce a latent factor model for tabular data errors that is simple to implement and can effectively model realistic error dependencies. Error sampling is decoupled from error types, which allows for simple extensions with more error types or sampling mechanisms. Using established benchmarks, we evaluate our model in two application scenarios, data cleaning and tabular ML tasks. In a comprehensive suite of experiments we demonstrate the impact of realistic error models on data cleaning benchmarks. Our results also show that a simple generative error model captures a wide range of error mechanisms and offers a convenient formalization of data perturbations to improve the generalizability, robustness and reproducibility of data cleaning research.

This talk will focus on our experience in managing the complexity of Sibyl, a large scale machine learning system that is widely used within Google. We believe that a large fraction of the challenges faced by Sibyl are inherent to large scale production machine learning and that other production systems are likely to encounter them as well [1]. Thus, these challenges present interesting opportunities for future research. The Sibyl system is complex for a number of reasons. We have learnt that a complete end-to-end machine learning solution has to have subsystems to address a variety of different needs: data ingestion, data analysis, data verification, experimentation, model analysis, model serving, configuration, data transformations, support for different kinds of loss functions and modeling, machine learning algorithm implementations, etc. Machine learning algorithms themselves constitute a relatively small fraction of the overall system.Each subsystem consists of a number of distinct components to support the variety of product needs. For example, Sibyl supports more than 5 different model serving systems, each with its own idiosyncrasies and challenges. In addition, Sibyl configuration contains more lines of code than the core Sibyl learner itself. Finally existing solutions for some of the challenges don't feel adequate and we believe these challenges present opportunities for future research.Though the overall system is complex, our users need to be able to deploy solutions quickly. This is because a machine learning deployment is typically an iterative process of model improvements. At each iteration, our users experiment with new features, find those that improve the model's prediction capability, and then launch a new model with those improved features. A user may go through 10 or more such productive launches. Not only is speed of iteration crucial to our users, but they are often willing to sacrifice the improved prediction quality of a high quality but cumbersome system for the speed of iteration of a lower quality but nimble system.In this talk I will give an example of how simplification drives systems design and sometimes the design of novel algorithms.

In recent years, many incidents have been reported where machine learning models exhibited discrimination among people based on race, sex, age, etc. Research has been conducted to measure and mitigate unfairness in machine learning models. For a machine learning task, it is a common practice to build a pipeline that includes an ordered set of data preprocessing stages followed by a classifier. However, most of the research on fairness has considered a single classifier based prediction task. What are the fairness impacts of the preprocessing stages in machine learning pipeline? Furthermore, studies showed that often the root cause of unfairness is ingrained in the data itself, rather than the model. But no research has been conducted to measure the unfairness caused by a specific transformation made in the data preprocessing stage. In this paper, we introduced the causal method of fairness to reason about the fairness impact of data preprocessing stages in ML pipeline. We leveraged existing metrics to define the fairness measures of the stages. Then we conducted a detailed fairness evaluation of the preprocessing stages in 37 pipelines collected from three different sources. Our results show that certain data transformers are causing the model to exhibit unfairness. We identified a number of fairness patterns in several categories of data transformers. Finally, we showed how the local fairness of a preprocessing stage composes in the global fairness of the pipeline. We used the fairness composition to choose appropriate downstream transformer that mitigates unfairness in the machine learning pipeline.

Deep-layers-assisted machine learning for accurate image segmentation of complex materials

To understand students’ learning behaviors, this study uses machine learning technologies to analyze the data of interactive learning environments, and then predicts students’ learning outcomes. This study adopted a variety of machine learning classification methods, quizzes, and programming system logs, found that students’ learning characteristics were correlated with their learning performance when they encountered similar programming practice. In this study, we used random forest (RF), support vector machine (SVM), logistic regression (LR), and neural network (NN) algorithms to predict whether students would submit on time for the course. Among them, the NN algorithm showed the best prediction results. Education-related data can be predicted by machine learning techniques, and different machine learning models with different hyperparameters can be used to obtain better results.

The UK financial sector increasingly employs machine learning techniques to enhance revenue and understand customer behaviour. In this study, we develop a machine learning workflow for high classification accuracy and improved prediction confidence using a binary classification approach on a publicly available dataset from a Portuguese financial institution as a proof of concept. Our methodology includes data analysis, transformation, training, and testing machine learning classifiers such as Naïve Bayes, Decision Trees, Random Forests, Support Vector Machines, Logistic Regression, Artificial Neural Networks, AdaBoost, and Gradient Descent. We use stratified k-folding (k=5) cross-validation and assemble the top-performing classifiers into a decision-making committee, resulting in over 92% accuracy with two-thirds voting confidence. The workflow is simple, adaptable, and suitable for UK banks, demonstrating the potential for practical implementation and data privacy. Future work will extend our approach to UK banks, reformulate the problem as a multi-class classification, and introduce pre-training automated steps for data analysis and transformation.

Sustainable urban development in the era of energy and digital transformation is crucial from a societal perspective. Utilizing modern techniques for analyzing large datasets, including machine learning and artificial intelligence, enables a deeper understanding of historical data and the efficient prediction of future events based on data from IoT sensors. This study conducted a multidimensional historical analysis of air pollution to investigate the impacts of energy transformation and environmental policy and to determine the long-term environmental implications of certain actions. Additionally, machine learning (ML) techniques were employed for air pollution prediction, taking spatial factors into account. By utilizing multiple low-cost air sensors categorized as IoT devices, this study incorporated data from various locations and assessed the influence of neighboring sensors on predictions. Different ML approaches were analyzed, including regression models, deep neural networks, and ensemble learning. The possibility of implementing such predictions in publicly accessible IT mobile systems was explored. The research was conducted in Krakow, Poland, a UNESCO-listed city that has had long struggle with air pollution. Krakow is also at the forefront of implementing policies to prohibit the use of solid fuels for heating and establishing clean transport zones. The research showed that population growth within the city does not have a negative impact on PMx concentrations, and transitioning from coal-based to sustainable energy sources emerges as the primary factor in improving air quality, especially for PMx, while the impact of transportation remains less relevant. The best results for predicting rare smog events can be achieved using linear ML models. Implementing actions based on this research can significantly contribute to building a smart city that takes into account the impact of air pollution on quality of life.

Rationale Multimodal clinical audio data, including speech and breath sounds, play a critical role in assessing respiratory function in disorders affecting the thoracic cage, including neurodegenerative, infectious, and respiratory disorders. Machine learning (ML) has shown promise in identifying patterns to enhance predictive accuracy in these disorders due to its data processing power with chronic respiratory conditions that disrupt airflow, such as Chronic Obstructive Pulmonary Disease (COPD) and Interstitial Lung Disease (ILD), being notable examples. This study investigates the feasibility of using an ML model to predict clinical parameters including pulmonary function, smoking status, and disease presence by analyzing breath and vocal characteristics. Methods Two-hundred patients with various respiratory and thoracic pathologies were recruited via convenience sampling from a university medical center. Data were collected using digital stethoscopes and microphones to record breath sounds at eight anatomical locations along with vocal samples during a variety of vocal tasks (Figure 1). Audio data were converted into Mel-spectrograms and used to train a convolutional neural network (CNN) model to predict patient outcomes, including gender, smoking status, COPD, and ILD. The dataset was divided into a 90:10 training-to-validation ratio. Fig. 1. Representation of study workflow showing data collection, data transformation, and subsequent model training. Results In predicting smoking status, breath recordings achieved 80% accuracy, 95% specificity, 63% sensitivity, and an AUC of 0.82 compared to 65%, 69%, 57%, and 0.58 for the voice-based model. In COPD classification, breath data yielded 75% accuracy, 62% specificity, and 80% sensitivity in contrast to 62.5%, 80%, and 57% for the vocal model. In ILD classification, breath data yielded 65% accuracy, 58% specificity, and 75% sensitivity in contrast to 65%, 77%, and 55% for the vocal model. Additionally, in predicting pulmonary function via Forced Vital Capacity (FVC) and the FEV1/FVC ratio, breath sounds had lower mean squared error (MSE) values relative to voice models (0.7 vs. 10.2, and 111.6 vs. 2308.6), demonstrating superior performance. Conclusion Breath sounds offer significant predictive value in identifying pulmonary conditions and estimating functional metrics, outperforming vocal and cough data in most tasks. Given the low cost of obtaining clinical audio data, uncovering reliable biomarkers in audio patterns via ML methods may assist in diagnosis, risk stratification, and identification of patients in whom preventative care may be beneficial. Future work will require larger datasets to improve model robustness and explore more complex biomarker integration, such as combining breath and voice data, for enhanced clinical prediction.

Mathematics serves as the foundational backbone of “artificial intelligence (AI) and machine learning (ML), providing the essential” tools and frameworks for developing sophisticated algorithms and models. the pivotal role of various mathematical disciplines, including linear algebra, calculus, probability theory, and optimization, in advancing AI and ML technologies. We begin by examining how linear algebra facilitates the manipulation and transformation of high-dimensional data, which is crucial for “techniques such as principal component analysis (PCA) and singular value decomposition (SVD)”. Next, we delve into the applications of calculus in training neural networks through gradient-based optimization methods, highlighting the importance of differentiation and integration in backpropagation and loss function minimization. the role of probability theory in handling uncertainty and making predictions, emphasizing its application in Bayesian networks, Markov decision processes, and probabilistic graphical models. Additionally, we discuss optimization techniques, both convex and non-convex, that are fundamental to finding optimal solutions in machine learning tasks, including support vector machines (SVMs) and deep learning architectures.

Privacy preserving in machine learning is a crucial issue in industry informatics since data used for training in industries usually contain sensitive information. Existing differentially private machine learning algorithms have not considered the impact of data correlation, which may lead to more privacy leakage than expected in industrial applications. For example, data collected for traffic monitoring may contain some correlated records due to temporal correlation or user correlation. To fill this gap, in this article, we propose a correlation reduction scheme with differentially private feature selection considering the issue of privacy loss when data have correlation in machine learning tasks. The proposed scheme involves five steps with the goal of managing the extent of data correlation, preserving the privacy, and supporting accuracy in the prediction results. In this way, the impact of data correlation is relieved with the proposed feature selection scheme, and moreover the privacy issue of data correlation in learning is guaranteed. The proposed method can be widely used in machine learning algorithms, which provide services in industrial areas. Experiments show that the proposed scheme can produce better prediction results with machine learning tasks and fewer mean square errors for data queries compared to existing schemes.

Parkinson’s disease is a neurodegenerative disease that is associated with genetic and environmental factors. However, the genes causing this degeneration have not been determined, and no reported cure exists for this disease. Recently, studies have been conducted to classify diseases with RNA-seq data using machine learning, and accurate diagnosis of diseases using machine learning is becoming an important task. In this study, we focus on how various feature selection methods can improve the performance of machine learning for accurate diagnosis of Parkinson’s disease. In addition, we analyzed the performance metrics and computational costs of running the model with and without various feature selection methods. Experiments were conducted using RNA sequencing—a technique that analyzes the transcription profiling of organisms using next-generation sequencing. Genetic algorithms (GA), information gain (IG), and wolf search algorithm (WSA) were employed as feature selection methods. Machine learning algorithms—extreme gradient boosting (XGBoost), deep neural network (DNN), support vector machine (SVM), and decision tree (DT)—were used as classifiers. Further, the model was evaluated using performance indicators, such as accuracy, precision, recall, F1 score, and receiver operating characteristic (ROC) curve. For XGBoost and DNN, feature selection methods based on GA, IG, and WSA improved the performance of machine learning by 10.00% and 38.18%, respectively. For SVM and DT, performance was improved by 0.91% and 7.27%, respectively, with feature selection methods based on IG and WSA. The results demonstrate that various feature selection methods improve the performance of machine learning when classifying Parkinson’s disease using RNA-seq data.

Data preprocessing (transformation) plays an important role in data mining and machine learning. In this study, we investigate the effect of four different preprocessing methods to fault-proneness prediction using nine datasets from NASA Metrics Data Programs (MDP) and ten classification algorithms. Our experiments indicate that log transformation rarely improves classification performance, but discretization affects the performance of many different algorithms. The impact of different transformations differs. Random forest algorithm, for example, performs better with original and log transformed data set. Boosting and NaiveBayes perform significantly better with discretized data. We conclude that no general benefit can be expected from data transformations. Instead, selected transformation techniques are recommended to boost the performance of specific classification algorithms.

We aimed to evaluate the diagnostic performance of neuroimaging-based machine learning (ML) models for non-invasive prediction of H3K27M mutations in diffuse midline gliomas (DMG). Following PRISMA-DTA guidelines, we searched four databases up to May 2025 to identify eligible studies evaluating neuroimaging-based ML for predicting H3K27M mutations in DMG. Study quality was assessed using PROBAST+AI and GRADE. Bivariate random-effects models were used to pool sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Sixteen studies were included, comprising 2,357 patients in internal validation cohorts and 1,792 patients in external validation cohorts. MRI-based ML models showed strong diagnostic performance in internal validation, with pooled sensitivity of 0.86 (95% CI: 0.79-0.91), specificity of 0.82 (95% CI: 0.75-0.87), and AUC of 0.91 (95% CI: 0.88-0.93). For PET/CT-based ML models, pooled sensitivity was 0.58 (95% CI: 0.44-0.71) and pooled specificity was 0.65 (95% CI: 0.46-0.81), with an AUC of 0.61 (95% CI: 0.57-0.66). MRI-based ML showed significantly higher sensitivity (Z = 3.71; P < 0.01) and AUC (Z = 11.42; P < 0.01) than PET/CT-based ML. Subgroup analysis indicated that MRI-based deep learning (DL) models outperformed conventional machine learning (cML) algorithms (P = 0.01). Models using DNA sequencing as the reference standard showed higher specificity than those using immunohistochemistry (P < 0.001). MRI-based ML demonstrates high accuracy and generalizability for non-invasive H3K27M prediction in DMG, seemingly outperforming current PET/CT-based ML. The adoption of DL architectures and DNA sequencing as the reference standard may further improve performance, supporting the clinical utility of MRI-based ML for molecular stratification. https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=1078673, identifier CRD420251078673.

As artificial intelligence techniques become more sophisticated, we anticipate that robots collaborating with humans will develop their own intentions, leading to potential conflicts in interaction. This development calls for advanced conflict resolution strategies in physical human–robot interaction (pHRI), a key focus of our research. We use a machine learning (ML) classifier to detect conflicts during co-manipulation tasks to adapt the robot’s behavior accordingly using an admittance controller. In our approach, we focus on two groups of interactions, namely “harmonious” and “conflicting,” corresponding respectively to the cases of the human and the robot working in harmony to transport an object when they aim for the same target, and human and robot are in conflict when human changes the manipulation plan, e.g. due to a change in the direction of movement or parking location of the object. Co-manipulation scenarios were designed to investigate the efficacy of the proposed ML approach, involving 20 participants. Task performance achieved by the ML approach was compared against three alternative approaches: (a) a rule-based (RB) Approach, where interaction behaviors were rule-derived from statistical distributions of haptic features; (b) an unyielding robot that is proactive during harmonious interactions but does not resolve conflicts otherwise, and (c) a passive robot which always follows the human partner. This mode of cooperation is known as “hand guidance” in pHRI literature and is frequently used in industrial settings for so-called “teaching” a trajectory to a collaborative robot. The results show that the proposed ML approach is superior to the others in task performance. However, a detailed questionnaire administered after the experiments, which contains several metrics, covering a spectrum of dimensions to measure the subjective opinion of the participants, reveals that the most preferred mode of interaction with the robot is surprisingly passive. This preference indicates a strong inclination toward an interaction mode that gives more control to humans and offers less demanding interaction, even if it is not the most efficient in task performance. Hence, there is a clear trade-off between task performance and the preferred mode of interaction of humans with a robot, and a well-balanced approach is necessary for designing effective pHRI systems in the future.