Transformer Autoencoder for rregular time series

Transformers are advanced Neural Network architectures useful in Data Analysis, not only for LLMs. 

I recently published a paper that presents a method based on a Transformer Autoencoder architecture for analyzing sparse and irregular time series. The autoencoder utilizes a local attention mechanism to identify similarities and anomalies within the complex structure of irregular datasets. It demonstrates notable clustering capabilities for samples that share similar characteristics.

The method is useful in risk estimation. The application presented in this work focuses on detecting accounts at higher risk of non-technical losses in electrical power systems, mainly electricity theft, using real-world consumption data from Greece.

The proposed framework combines classical data cleaning procedures with a state-of-the-art transformer architecture, exploiting the local attention mechanism to capture meaningful short-range temporal patterns without requiring interpolation of missing data.

The paper is available here. Also a read only version and a preprint.

Java's birthday

Peer review process at a glance

 

A guided framework for LLM-based risk estimation

Data Science is going through a transformation. By leveraging the impressive reasoning capabilities of generative AI, we can overcome manual data analysis and perform data analysis faster than ever before.

LLMs are moving rapidly from chat windows into operational pipelines. This shift is not risk-free as they are not yet trustworthy enough. LLMs suffer from hallucinations when they fabricate facts to generate answers they don't know. There is also the alignment problem, where the models misinterpret the task they were asked to perform.

In my latest paper titled “Towards automated data analysis: A guided framework for LLM-based risk estimation”, I discuss these issues and present a framework that uses LLMs to perform risk analysis, supervised by a Human-in-the-Loop. Rather than relying on a risky single-prompt approach, I propose a four-stage structured framework, which allows a human supervisor to verify the integrity and accuracy of each stage before the model moves on to the next stage.

The four stages:
1) The model identifies entities and relations of the given dataset and suggests clustering techniques.
2) Generates the code for implementing the suggested techniques.
3) The user, or an agent, executes the code.
4) The model analyzes the produced results and generates a comprehensive report.

To demonstrate the viability of this approach, the paper includes a proof of concept applying the framework to a real-world problem. We examine risk estimation of non-technical losses on power grids. The results show that within this framework, the LLM does not act just as a simple chat-bot assistant, but as a system capable of producing full, reliable data analysis reports with some degree of automation.

I just completed my PhD !!!

 Last Thursday, I successfully defended my doctoral thesis and completed my Ph.D. from the University of Macedonia!

This 5 year journey was the most challenging project I have undertaken so far and the sense of accomplishment is profound.

I am thankful to my advisor Panagiotis Papadimitriou for his invaluable guidance and fruitful co-operation.

My thesis titled "AI-assisted resource orchestration for network services and distributed cloud-native applications" develops methods for leveraging Artificial Intelligence for the orchestration of network services and cloud-native applications.

You can access the full dissertation here: 

http://dspace.lib.uom.gr/handle/2159/33955