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.

We investigate the enhancement of energy efficiency in multi-band optical networks through the deployment of inter-band all-optical wavelength converters (AO-WCs). To address the ever-increasing traffic demand, multi-band wavelength-division multiplexing transmission technologies are widely deployed to expand network capacity. However, in all-optical networks, the efficiency of wavelength resource utilization faces limitations due to wavelength continuity constraints. To cope with this issue, AO-WCs are intensively investigated for energy-efficient all-optical networking. AO-WCs significantly reduce energy consumption compared to current optical-electrical-optical wavelength converters (OEO-WCs). However, a typical inter-band AO-WC merely converts an input C -band (or L-band) wavelength to the mirror-image position of the output L-band (or C-band) wavelength. This limitation may impact the effectiveness of AO-WCs in terms of energy efficiency and network resource utilization. We thus quantitatively evaluate the performance of multi-band optical networks with inter-band AO-WCs. Our evaluation results demonstrate that inter-band AO-WCs can accommodate 20% more traffic demand compared to scenarios without any wavelength converters, while significantly improving energy efficiency compared to OEO-WCs.

There is a strong demand for creating primary/backup optical path pairs between a source and a destination node in order to continue a service when the primary path has some disorder. We also have to consider SRLG (shared risk link group)-disjoint primary/backup path settings for achieving robust path protection against possible network problems on primary paths. To deal with these issues, in this study, we use the k-SPF (shortest path first) algorithm to determine primary/backup path routes between a specified source/destination node pair. We demonstrate that the proposed algorithm can find$k$SRLG-disjoint primary/backup route pairs with much higher probability and with faster speed compared with other algorithms. In addition, the proposed algorithm can find primary/backup routes with nearly minimum route costs.

The fabric architecture, which can realize massive packet forwarding even with small switches/routers, has been widely used as a network infrastructure in data centers in order to achieve high scalability and low cost. In carrier networks, large scale routers have been adapted as core routers due to the requirement of high density for accommodating a large number of users. Furthermore, the redundant configuration of these routers can be used for high reliability. However, these traditional approaches result in a significant cost increase. The fabric architecture has the potential to solve this problem. In this paper, we propose a transformation from the router based architecture to the fabric based architecture. We also evaluate the reliability of these two architectures.

In the 5G era, it is expected that various services will be provided on network slices, which are virtualized network resources. Unlike the uniform service level provided in the current Internet, a network slice has to support multiple service levels for various types of services such as IP telephony, Internet-access and IoT/M2M. Because their requirements are quite different. IP telephony and Internet-access services require high reliability due to their essential role in communication. However, some IoT/M2M services do not necessarily require high reliability since they can tolerate temporary disconnection of networks due to their asynchronous communications. Instead of reliability, they require low-cost connectivity. Therefore, future networks have to support multi-level reliability required by services on a single physical network. In this paper, we propose a multi-level reliability architecture in optical cross-connect based metro networks to satisfy broad service requirements.