Reading Guide: Advances and Trends in Artificial Intelligence (IEA/AIE 2019)

Advances and Trends in Artificial Intelligence. From Theory to Practice is the proceedings of IEA/AIE 2019, the 32nd International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, held in Graz, Austria, on July 9 to 11, 2019. Published by Springer as volume 11606 of Lecture Notes in Computer Science (in the Lecture Notes in Artificial Intelligence sub-series), it was edited by Franz Wotawa, Gerhard Friedrich, Ingo Pill, Roxane Koitz-Hristov and Moonis Ali. The volume contains 41 full papers and 32 short papers, selected from 151 submissions, across roughly 865 pages. This guide helps you navigate it by theme.

What this book is (and who it's for)

IEA/AIE is a long-running conference about applied intelligent systems: taking techniques from AI research and making them work on real problems in engineering, industry, automation and robotics, business and finance, medicine and bioinformatics, cyberspace and human-machine interaction. Springer notes that the 2019 edition gave special attention to automated driving, autonomous systems, and how to verify and validate them, a fitting focus for a host city with a strong automotive and testing community.

The volume will be most useful to:

• Graduate students and researchers who want a broad view of applied AI just before the large-language-model era.
• Engineers in industry working on diagnosis, forecasting, robotics or decision support who want concrete case studies.
• Anyone studying AI safety and dependability, thanks to the autonomous-systems and fault-diagnosis material.

You should be comfortable with core AI and ML concepts: search and reasoning, classification and regression, neural networks and basic probability. Each paper is self-contained, so no single chapter is a prerequisite for another.

Before you start

• Know the format. Full papers give complete methods and evaluations; short papers are compact and often describe early results. Read both, but weight your conclusions accordingly.
• Remember the date. This is 2019 research. Treat it as a strong record of methods and problem framings, and check newer literature before adopting a specific technique.
• Keep a comparison sheet with columns for domain, AI technique, data, evaluation and limitations. Across 73 papers, patterns show up quickly.

How to read it

The proceedings are organized into eleven topical sections. Here is a roadmap with suggested entry points for different readers.

AI for Estimation and Prediction

Forecasting and prediction problems: weather forecasting for holiday planning, a taxonomy of event-prediction methods, filling gaps in rainfall and runoff data with a Gaussian-mixture-based nearest-neighbor approach, predicting the listing status of companies, estimating query execution times on parallel databases, and active learning in probabilistic programs for travel-time estimation. Start with the event-prediction taxonomy for a conceptual overview.

Applied Neural Networks

Fish species classification with squeeze-and-excitation CNNs, open-set text classification, deep learning for hydropower optimization, and lightweight residual networks for real-time head pose estimation. Good for seeing how network design trades accuracy against speed.

Autonomous Systems and Automated Driving

The conference's special focus. Papers include rule-based control for fail-operational systems, air-quality monitoring with drones, requirements engineering for intelligent parking, and neural control of robot motion among obstacles. Start here if you care about safety-critical autonomy; the fail-operational paper sets out the core concern clearly.

Fault Detection and Diagnosis

A traditional strength of this community, closely linked to model-based reasoning. Read this alongside the autonomous-systems section: together they show how intelligent systems detect, explain and recover from failure.

Knowledge Representation and Reasoning

The "theory" half of the subtitle. Recommended for readers who want to understand symbolic approaches that complement learned models.

Decision Support and Recommender Systems; Intelligent Information Storage and Retrieval

Systems that help people choose, search and organize information. A good entry point for readers with a business or information-systems background.

Data Science and Security; Natural Language Processing and Sentiment Analysis

Data-driven analytics, security applications, and language work from the pre-transformer-dominance period. Useful for seeing which problems were already well framed before today's large models.

Mobile and Autonomous Robotics; Intelligent Systems in Real-Life Applications

Deployment-oriented papers. Skim the abstracts and choose the domains closest to your own work.

Suggested first-pass plans: an autonomy and safety reader should take Autonomous Systems, then Fault Detection, then Knowledge Representation. A data scientist should take Estimation and Prediction, then Applied Neural Networks, then Data Science and Security. A business or IS reader should take Decision Support, then Information Retrieval, then Real-Life Applications. For anything else, read abstracts and conclusions only, and dive in where something clicks.

Key ideas to take away

• Applied AI is a mix of methods. Symbolic reasoning, probabilistic models, classic ML and deep learning all appear side by side, chosen to fit the problem.
• Verification matters as much as performance for autonomous and safety-critical systems.
• Diagnosis is a core AI capability. Knowing why a system fails is often more valuable than a small accuracy gain.
• Real-world data is messy. Imputation, open-set problems and concurrency effects recur throughout the volume.
• Deployment constraints shape design, as the real-time and parameter-reduced network papers show.

Practice and discussion questions

1. What does "fail-operational" mean, and how does it differ from "fail-safe"? Where would each be appropriate in a vehicle?
2. Pick one paper from Estimation and Prediction. How would you evaluate it on data from a different region or time period?
3. When would you choose a rule-based controller over a learned one in an autonomous system?
4. How might a diagnosis approach from this volume be combined with a modern deep-learning perception stack?
5. Choose an NLP or sentiment paper. Which parts would change if you rebuilt it with today's language models, and which would not?
6. What evidence would convince you that a short paper's early result is worth pursuing further?

Pair it with

• Model-based diagnosis and formal verification, to go deeper on the dependability theme.
• A modern survey of autonomous-driving perception and planning, to see how the field moved after 2019.
• Experimental methodology in ML, so you can judge the evaluation sections critically.

Pick a track and dive in

With 73 papers across eleven topics, this volume rewards readers who follow one track at a time. Choose the path that matches your goals and let the citations lead you on. Find this and related AI proceedings on the Books Plus Ai bookshelf.

Sources

• Springer: Advances and Trends in Artificial Intelligence. From Theory to Practice (IEA/AIE 2019)
• dblp: IEA/AIE 2019 proceedings listing