TASK Quarterly

Two architectures of neural networks in distance approximation

2024-12-04T11:24:13+01:00

In this research paper, we examine recurrent and linear neural networks to determine the relationship between the
amount of data needed to achieve generalization and data dimensionality, as well as the relationship between data
dimensionality and the necessary computational complexity. To achieve this, we also explore the optimal topologies
for each network, discuss potential problems in their training, and propose solutions. In our experiments, the relation-
ship between the amount of data needed to achieve generalization and data dimensionality was linear for feed-forward
neural networks and exponential for recurrent ones. However, the required computational complexity appears to grow
exponentially with increasing dimensionality. We also compared the networks’ accuracy in both distance approxima-
tion and classification to the most popular alternative, siamese networks, which outperformed both linear and recurrent
networks in classification despite having lower accuracy in exact distance approximation.

Document coordination patterns

2025-06-10T14:31:20+02:00

Many organizations lack support for document workflow management, making it difficult to efficiently handle business processes. Implementing a workflow management service that encompasses all organizational tasks is both complex and costly. However, document workflows can be observed as structured sets of identifiable workflow patterns. These patterns have been extensively described in the literature based on real-world organizational processes. By adopting and integrating these patterns into document workflows, it is possible to enable documents to autonomously determine the processes they should execute. The paper explores the application of a well-known set of workflow patterns in email-based document workflows and outlines the necessary conditions for integrating workflow management directly into documents. Documents circulating within an organization can easily collect information about their workflow in the form of logs with minimal effort. By applying process mining techniques, it is possible to extract the actual processes they follow. The proposed in the paper canonical set of document workflow patterns enables a much faster implementation of a document management application compared to a top-down approach, where processes are theoretically defined first and then implemented.

A Survey on Privacy-Preserving Machine Learning Inference

2025-04-01T15:30:55+02:00

This paper examines methods to secure machine learning inference (ML inference) so that sensitive data remains private and proprietary models are protected during remote processing. We review several approaches—from cryptographic techniques like homomorphic encryption (HE) and secure multi-party computation (MPC), to hardware solutions such as trusted execution environments (TEEs), and complementary methods including differential privacy and split learning. Each method is analyzed in terms of security, efficiency, communication overhead, and scalability. Use cases in healthcare, finance, and education show how these techniques balance privacy with practical performance. We conclude by outlining open challenges and future directions for building robust, efficient privacy-preserving ML inference systems.