Auto Feature Engineering (AFE) plays a crucial role in developing practical machine learning pipelines by automating the transformation of raw data into meaningful features that enhance model performance. By generating features in a data-driven manner, AFE enables the discovery of important features that may not be apparent through human experience or intuition. On the other hand, since AFE generates features based on data, there is a risk that these features may be overly adapted to the data, making it essential to assess their reliability appropriately. Unfortunately, because most AFE problems are formulated as combinatorial search problems and solved by heuristic algorithms, it has been challenging to theoretically quantify the reliability of generated features. To address this issue, we propose a new statistical test for generated features by AFE algorithms based on a framework called selective inference. As a proof of concept, we consider a simple class of tree search-based heuristic AFE algorithms, and consider the problem of testing the generated features when they are used in a linear model. The proposed test can quantify the statistical significance of the generated features in the form of $p$-values, enabling theoretically guaranteed control of the risk of false findings.

A data analysis pipeline is a structured sequence of steps that transforms raw data into meaningful insights by integrating various analysis algorithms. In this paper, we propose a novel statistical test to assess the significance of data analysis pipelines. Our approach enables the systematic development of valid statistical tests applicable to any feature selection pipeline composed of predefined components. We develop this framework based on selective inference, a statistical technique that has recently gained attention for data-driven hypotheses. As a proof of concept, we focus on feature selection pipelines for linear models, composed of three missing value imputation algorithms, three outlier detection algorithms, and three feature selection algorithms. We theoretically prove that our statistical test can control the probability of false positive feature selection at any desired level, and demonstrate its validity and effectiveness through experiments on synthetic and real data. Additionally, we present an implementation framework that facilitates testing across any configuration of these feature selection pipelines without extra implementation costs.

In real-world applications, anomaly detection (AD) often operates without access to anomalous data, necessitating semi-supervised methods that rely solely on normal data. Among these methods, deep $k$-nearest neighbor (deep $k$NN) AD stands out for its interpretability and flexibility, leveraging distance-based scoring in deep latent spaces. Despite its strong performance, deep $k$NN lacks a mechanism to quantify uncertainty—an essential feature for critical applications such as industrial inspection. To address this limitation, we propose a statistical framework that quantifies the significance of detected anomalies in the form of $p$-values, thereby enabling control over false positive rates at a user-specified significance level (e.g.,0.05). A central challenge lies in managing selection bias, which we tackle using Selective Inference—a principled method for conducting inference conditioned on data-driven selections. We evaluate our method on diverse datasets and demonstrate that it provides reliable AD well-suited for industrial use cases.

The Vision Transformer (ViT) demonstrates exceptional performance in various computer vision tasks. Attention is crucial for ViT to capture complex wide-ranging relationships among image patches, allowing the model to weigh the importance of image patches and aiding our understanding of the decision-making process. However, when utilizing the attention of ViT as evidence in high-stakes decision-making tasks such as medical diagnostics, a challenge arises due to the potential of attention mechanisms erroneously focusing on irrelevant regions. In this study, we propose a statistical test for ViT's attentions, enabling us to use the attentions as reliable quantitative evidence indicators for ViT's decision-making with a rigorously controlled error rate. Using the framework called selective inference, we quantify the statistical significance of attentions in the form of p-values, which enables the theoretically grounded quantification of the false positive detection probability of attentions. We demonstrate the validity and the effectiveness of the proposed method through numerical experiments and applications to brain image diagnoses.

In this study, we investigate the quantification of the statistical reliability of detected change points (CPs) in time series using a recurrent neural network (RNN). Thanks to its flexibility, RNN holds the potential to effectively identify CPs in time series characterized by complex dynamics. However, there is an increased risk of erroneously detecting random noise fluctuations as CPs. The primary goal of this study is to rigorously control the risk of false detections by providing theoretically valid p-values to the CPs detected by RNN. To achieve this, we introduce a novel method based on the framework of selective inference (SI). SI enables valid inferences by conditioning on the event of hypothesis selection, thus mitigating bias from generating and testing hypotheses on the same data. In this study, we apply an SI framework to RNN-based CP detection, where characterizing the complex process of RNN selecting CPs is our main technical challenge. We demonstrate the validity and effectiveness of the proposed method through artificial and real data experiments.

Carbon emissions from peatland significantly affect climate change. We adopted SAR interferometry technique in order to estimate the amount of carbon emissions from peatland receiving development pressure in Borneo. As a result, ground subsidence of 15.0 cm yr-1on average and 31.0 cm yr-1at maximum were observed. The carbon emissions were estimated at 39.7 tC ha-1yr-1. We will compare the results with in-situ subsidence data and CO2flux data in the future,

The temporal variations (diurnal and annual) in arboreal (εTree) and bare soil (εSoil) dielectric constants and their correlation with precipitation were examined for several trees in Japan. A significant (1 σ (standard deviation) and 2 σ) εTreeincrease is observed after rainfall at 89.8% and 90.5% probability. However, rainfall does not always induce significant εTreeincreases. Rainfall of more than 5 mm/day can induce 1 σ εTreeTree increase at a 59.6% probability. In order to examine whether the increase in εTree affects the L-band σ0variation in a forest, the four-year temporal variation of the L-band backscattering coefficient (σ0) was estimated from observations by the Advanced Land Observing Satellite Phased Array type L-band Synthetic Aperture Radar. Observed maximum absolute deviations from the mean over the forest area were 1.0 and 1.2 dB for σHH0and σHV0, respectively, and 4.0 and 3.0 dB over open land. σ0and rainfall correlations show that εTreeand σForest0are proportional to precipitation integrated over seven or eight days; εSoiland σOpen land0are proportional to precipitation integrated over three days. This finding indicates that εTreevariations influence σForest areas0. A stronger correlation between σHV0and precipitation is observed in several sites with low σHV0, where less biomass is expected, and several sites with high σHV0, where more biomass is expected. A weaker correlation between σHV0and precipitation is observed for several sites with high σHV0. These differences may be explained by the different contributions of double bounce scattering and potential transpiration, which is a measure of the ability of the atmosphere to remove water from the surface through the processes of transpiration. The two other results were as follows: 1) The functional relation between aboveground biomass and σ0showed dependence on precipitation data, this being an effect connected with seasonal changes of the εTree. This experiment reinforces the fact that the dry season is preferable for retrieval of woody biomass from inversion of the functional dependence of SAR backscatter and for avoiding the influence of rainfall. 2) The complex dielectric constant for a tree trunk, which is measured between 0.2 and 6 GHz, indicates that free water is dominant in the measured tree.

Aboveground (AG)-biomass was estimated from a field biomass collection and LiDAR observations for a natural forest in Indonesia. The derived AG-biomass data were plotted against full polarimetric parameters calculated from Polarimetric and Interferometric Airborne Synthetic Aperture Radar L2 (Pi-SAR-L2) and PALSAR data. The α°-AG-biomass curve shows saturation by around 100 tons/ha, while for entropy, correlation is observed up to 200 tons/ha. The largest coefficient of determination (R2= 0.2348) was observed for the range with AG-biomass of more than 100 tons/ha for the relation between AG-biomass and entropy. The α°HV-biomass plot derived from PALSAR data shows smaller variance in the dry season than in other seasons, indicating that dry season data is preferable for a more accurate estimate of AG-biomass.

The Japan Aerospace Exploration Agency (JAXA) has produced the world's first 10m resolution L-band SAR global mosaic datasets. These data sets were generated to monitor forest changes from the 1990s to present. SRTM-3 (90m resolution) DEM was used to correct the terrain-induced SAR intensity variations and the ortho-rectification. Both corrections were applied for geometric and radiometric calibration purposes. The data sets are useful to monitor the temporal forest cover and forest change, and were used to derive forest/non-forest information.