The aim of this conference is to bring together European academics, young researchers, students and industrial practitioners to discuss the application of Artificial Intelligence in Industry and Finance.

The 1st COST Conference on this topic was held on September 15, 2016, the 2nd COST Conference was held on September 7, 2017, and the 3rd COST Conference was held on September 6, 2018.

All lectures are open to the public. Registration is now open.

• Artificial Intelligence in Finance: Artificial Intelligence and Fintech challenges for the European banking and insurance industry
• Artificial Intelligence in Industry: Artificial Intelligence challenges for European companies from the mechanical and electrical industry, but also life sciences.
• Big Data and Automation in Finance: This session is devoted to the use of Big Data and Fintech technologies in the ongoing process of automation in the financial industry.
• Ethical Questions in Artificial Intelligence (co-hosted by the Swiss Alliance for Data-Intensive Services): Issues arising in AI applications such as trust, explainability, neutrality, responsability, moral consequences of algorithmic decisions.
  

In September 2017 and 2018, we have had around 200 participants, both from Academia and Industry. The latest installment of the COST conference also saw a large number of international guests and speakers, travelling to Switzerland from destinations such as the UK, Germany, the United States and Bulgaria.

The largest proportion of participants will come from the industry complemented by a significant number of academic researchers. This mirrors our unique approach of connecting the academic world to their respective fields of application, putting new exciting concepts to work in industrial frameworks, where they can open up new opportunities.

 A Brief Biography

Dr.-Ing. Carlos Jiménez Härtel is a strategy advisor to private companies and public-sector organizations and holds a number of national and international board assignments. His operational career in industry spans about two decades in a variety of senior leadership roles, most recently as CTO & Chief Innovation Officer for GE Europe and previously as President & CEO for GE Germany. Carlos Härtel has extensive experience in industrial R&D, management of technology transfer and innovation, and the assessment of potential and market readiness of advanced technologies. He studied Aerospace Engineering at RWTH Aachen and TU Munich where he also received his doctorate. In addition, he spent several years in research at the German Aerospace Center (DLR) and at ETH Zurich, where he qualified as University Lecturer in 1999. Carlos Härtel is past President of the European Industrial Research Management Association (EIRMA).

Minds & Machines – Towards the Digital Industrial Company

Advanced digital technologies have become commonplace in many parts of daily life, but only recently started to make noticeable inroads in the industrial sector. Summarized under headlines like “Industry 4.0” or “Factory of the Future”, the suite of technologies relevant to industrial operations includes artificial intelligence and augmented reality, collaborating robots and 3D printing, or cyber-physical systems in general. The talk will give an overview of how industrial companies employ digital technologies today – and in particular AI – on various operational levels in order to increase productivity and improve business outcomes.

A Brief Biography

Saeed Amen is the founder of Cuemacro. Over the past fifteen years, Saeed Amen has developed systematic trading strategies at major investment banks including Lehman Brothers and Nomura. He is also the author of Trading Thalesians: What the ancient world can teach us about trading today (Palgrave Macmillan) and is the coauthor of The Book of Alternative Data (Wiley), due in 2020. Through Cuemacro, he now consults and publishes research for clients in the area of systematic trading. He has developed many Python libraries including finmarketpy and tcapy for transaction cost analysis. His clients have included major quant funds and data companies such as Bloomberg. He has presented his work at many conferences and institutions which include the ECB, IMF, Bank of England and Federal Reserve Board. He is also a co-founder of the Thalesians.

An introduction to natural language processing

In this presentation, we shall introduce the topic of NLP. We shall give a brief explanation of the various techniques used, such as word embeddings and topic modelling. We also discuss some of the Python tools that can be used to parse text and perform NLP tasks. We later give several use cases for finance such as the parsing of Fed communications to give an insight into bond yield moves.

A Brief Biography

Prof. Dr. Regina Betz is Head of the Center for Energy and the Environment at the School of Management and Law of the Zurich University of Applied Sciences (ZHAW). Before joining ZHAW she was Joint Director at the Centre for Energy and Environmental Markets (CEEM) at the UNSW Sydney. Regina’s teaching and research interests include energy and climate economics. She also contributes to master courses and consults to industry and government clients in these areas in Europe and Australia. She has been called to appear as an expert witness in climate change issues at various Federal government enquiries and is the co-president of the Swiss National Science Foundation (SNF) National Research Program on Sustainable Economy. Today her research is mainly focusing on the design of climate policies such as emissions trading schemes or energy efficiency obligations as well as electricity markets applying experimental economics or empirical methods.

Transfers of Kyoto units in the Swiss Emissions Trading Registry: A network analysis from 2007-2014

The Swiss emissions trading registry was opened in 2007 and since then companies – including foreign companies - or private persons were able to open an account and transfer units. Such an account enabled to hold and transfer different types of units issued from different jurisdictions including units based on the Kyoto Protocol or emission allowances (CHUs) generated by the Swiss government for compliance under the Swiss Emissions Trading System. The majority of transferred units in the Swiss registry were Kyoto units of the following type: 1) Certified Emissions Reductions (CERs), issued according to the rules of the Clean Development Mechanism (Article 12 des Kyoto-Protocol) and Emission Reduction Units (ERUs), issued according to the rules of Joint Implementation under Article 6 of the Kyoto Protocol.
The aim of this paper is to analyse the transfer flows of both national and international emissions allowances and international emission credits between different accounts in order to gain a picture of Switzerland's role in international emissions trading. We use network analysis to analyse the Swiss registry transfer data from 2007 to 2014. The analysis shows that, contrary to widespread belief, the transfer volume through the Swiss registry was substantial. Although transfers in Swiss emission allowances (CHUs) are negligible, the volumes of CERs and ERUs transferred via the Swiss registry are considerable. According to World Bank estimates the total trading volume of CERs in 2008 amounted to 1 billion, of which approximately 400 million transactions were recorded in the Swiss Registry. The type of actors in both submarkets vary considerably. The market of ERUs is dominated by commodity traders buying large volumes from Russia and the Ukraine, the CER market involves more players from different sectors and is less concentrated.

Aleksandra Chirkina

Aleksandra is a Senior Data Scientist at InCube working on designing and implementing machine learning solutions in the financial industry. She holds an MS degree in Computational Science from UvA (University of Amsterdam) and an MS degree in Statistics from the ETH. She has 2 years of experience in applying data analytics and numerical optimization in industry. 

Richard Jeroense

Richard leads the Digitalization and Process Excellence teams at Bank J. Safra Sarasin. They focus on leveraging technology and best practice process management methods to develop new client offerings and enhancing productivity. He holds a Bachelor in Industrial Engineering from the University of Eindhoven (NL), has 14 years’ experience in the Swiss private banking sector, and 10 years’ experience in the industrial and chemical sector across Europe. 

Affinity is in the AIRS: Personalized Investment Recommendations Delivered to Clients’ E-Services

Intelligent recommender for investment ideas is highly desirable for a private bank. Besides obvious advantages in providing relevant suggestions and saving the time of relationship managers, it can increase the usage of online banking platforms and improve clients’ engagement. As a part of the Bank J. Safra Sarasin digitization initiative, the AIRS project was launched. Its goal was to design an AI Recommender System for automatic generation of personalized investment ideas. InCube was selected as the fintech partner for the implementation of this system.

The project consisted of two phases: proof of concept and operationalization. Phase I was dedicated to an investigation whether a collaborative filtering approach is applicable for investment ideas at all. Financial data has certain characteristics a recommender should adapt to, such as implicit ratings and changing features of the items. After Phase I was successfully completed, the recommendations were evaluated by Relationship Managers. The goal of phase II was to put the recommendations into a shape deliverable to clients. This meant ensuring agreement with the regulations by imposing quality control, elaborating explanations, reacting to the clients’ feedback, as well as guaranteeing system robustness and reproducibility of results. 

A Brief Biography

Martin Hillebrand is a Senior Analyst at the European Stability Mechanism. Prior to joining the ESM, he worked as Quantitative Analyst in the Trading & Derivatives department of Sal. Oppenheim and as Risk Analyst at both Deutsche Bank and the German Finance Agency (Deutsche Finanzagentur). He holds a PhD in Mathematics from the University of Oldenburg. In 2008, he received the Professional Risk Managers International Association (PRMIA) “Award for New Frontiers in Risk Management” for an outstanding research paper. His current research focuses on fixed income markets and credit risk. In addition to his role at the ESM, Martin is a lecturer at the Frankfurt University of Applied Sciences.

Predicting Investor Behaviour in European Bond Markets. A Machine-Learning Approach

The European Rescue Fund ESM has, in its role as financial backstop of the Euro area, a specific interest in a comprehensive understanding of investor behaviour in order to ensure a stable and broad market access.

With numerous transaction data as well as market and macro variables, a learning machine has been trained that forecasts investor demand in syndicated transactions. Out-of-sample tests show already a decent predictive power which is intended to be further improved by intelligent methods of data enhancement.

A Brief Biography

Martin is the CEO and founder of Bulletproof AI. Bulletproof AI builds solution for security of machine learning and statistical techniques applied to credit risk scoring, fraud detection, anti-money laundering and biometric authentication, as well as other AI application domains in finance and security industry. He is also a veture partner with Credo Ventures and an active angel investor in security and enterprise technology fields. Martin has been active in AI & Security fields since 2003. Prior to his ccurrent position, Martin led the Cisco's Cognitive Threat Analytics (CTA) team. CTA was part of the Advanced Threat portfolio and provided advanced threat detection by analysis of network traffic for more than 25 million users worldwide. Prior to his Cisco role, he was the CEO & Founder of Cognitive Security, acquired by Cisco in 2013. Martin holds an engineering degree from Ecole Centrale Paris and a Ph.D. in AI from CTU in Prague.

AI vs. AI - Intelligent Attacks on Automated Financial Decisions

Increasing use of machine learning enables automation and immediate delivery of tasks that were previously tedious, long or even impossible. On the other hand, AI techniques are not magic and, besides other issues,  are susceptible to attacks, manipulation and unintentional bias. The attacks against AI can be broken down into three categories. Attacks on confidentiality extract the knowledge stored in the AI models. Evasion attacks use existing vulnerabilities in model definition and training to force unintentional decision of the model. The poisoning attacks are based on model manipulation by introducing carefully crafted training data to influence future model decisions. The current consensus in the research community appears to be that there is no universal solution to make models safe. Therefore, we propose a solution that protects the model from the outside by generative, AI-based threat modeling, intelligent security probing of the model, better training and continuous monitoring of production model behavior for signs of attack or bias.

A Brief Biography

Mario Wüthrich is Professor in the Department of Mathematics at ETH Zurich. He holds a PhD in Mathematics from ETH Zurich (1999). From 2000 to 2005, he held an actuarial position at Winterthur Insurance, Switzerland. He is fully qualified actuary of the Swiss Association of Actuaries, served on the board of the Swiss Association of Actuaries (2006-2018), and is Editor-in-Chief of ASTIN Bulletin.

Yes, we CANN!

We illustrate how we can smoothly transition from classical statistical models (like generalized linear models GLMs) to neural network architectures. In fact, this transition provides us with a natural blending of the data modeling culture with the algorithmic modeling culture (Leo Breiman, 2001), and it allows us to back-test classical statistical models with neural network features. We illustrate this approach on a car insurance data set.

A Brief Biography

Milos Cernak holds a PhD. degree in electrical engineering from Slovak University of Technology in Bratislava. From 2011 to 2018 he was is a senior engineer and associate researcher at Idiap Research Institute, Martigny, Switzerland. From 2008, he was a member of IBM Research in Prague. After graduating in 2005, he was a post-doc researcher at Institute EURECOM in France, and a principal researcher at Slovak Academy of Sciences. During his studies in 2001, he was also a visiting scientist at Iowa State University's Virtual Reality Application Centre, Ames, USA. Currently he is a principal software engineer at Logitech, CTO-AI group, in EPFL Innovation Park, Lausanne, Switzerland. His research interests include audio and speech signal processing, and applied machine learning. He is a Senior Member of the IEEE.

Applied Machine Learning at Logitech

This talk introduces the work of an AI group at Logitech, focusing on applied audio-visual machine learning. Three use cases are described: i) Keyword Spotting (KWS) systems that are in high demand nowadays as they enable a simple voice user interface to consumer electronics devices, ii) end-to-end accented speech recognition to cope with the absence of given accents in the training set where a voice interface becomes unusable for unseen accents, and iii) audio-visual source separation involving mono-channel audio where source contributions overlap both in time and frequency, applied to video conferencing.

A Brief Biography

Artificial Intelligence in Air Transport Industry: Airline schedule redesign in a case of disturbance

A Brief Biography

Michael Rupprecht is a Principal Solutions Architect at Amazon Web Services. He has spent his professional career in the  software, automotive, and financial services industries; always in IT architect or enterprise architect roles.

AWS Machine Learning – centerpiece of digital transformation

This presentation will provide an overview of the AWS AI Portfolio show how customers are using AI/ML to transform their business in particular how customers use AI/ML services to further personalize their business towards the customer.

A Brief Biography

Dr. Peter Staar joined the IBM Research - Zurich Laboratory in July of 2015 as a post-doctoral research fellow in the Foundations of Cognitive Solutions project. The Belgium-born scientist first came to IBM Research as a summer student in 2006.

Prior to joining IBM Research, Peter was a post-doctoral researcher in Theoretical Physics and PASC (Platform for Advanced Scientific Computing) at the Swiss Federal Institute of Technology (ETH) in Zurich, Switzerland.

He earned his PhD in Theoretical Physics and his M.Sc. degree in Physics at ETH Zurich in 2013 and 2009, respectively, and his B.S. degree in Physics (cum laude) from the Catholic University Leuven, Belgium.

Peter has twice been a finalist for the prestigious ACM Gordon Bell award, first in 2013 for his paper entitled "Taking a Quantum Leap in Time to Solution for Simulations of High-Tc Superconductors" and then in 2015 for his paper entitled "An Extreme-Scale Implicit Solver for Complex PDEs: Highly Heterogeneous Flow in Earth Mantle."

In his leisure time, Peter is an active sportsman and photographer who enjoys exploring exotic cultures.

AI assisted Scalable Knowledge Ingestion for Automated Discoveries

Over the past few decades, the amount of scientific articles and technical literature has increased exponentially in size. Consequently, there is a great need for systems that can ingest these documents at scale and make the contained knowledge discoverable. Unfortunately, both the format of these documents (e.g. the PDF format or bitmap images) as well as the presentation of the data (e.g. complex tables and figures) make the extraction of qualitative and quantitive data extremely challenging. In this talk, we will present our three pronged approach to this problem and show practical examples in the field of Material Science and Oil&Gas. We will start by introducing a scalable service [1] that is able to ingest documents at scale and exploits state-of-the art AI models to obtain very high accuracies. Next. we will show how the data contained in the ingested documents can be extracted using NLP methods. Finally, we will show how the extracted data can be efficiently queried using Knowledge Graphs and how one can obtain new insights from these graphs by applying advance analytics [2].

[1] https://www.researchgate.net/publication/325359423_Corpus_Conversion_Service_A_Machine_Learning_Platform_to_Ingest_Documents_at_Scale

[2] https://www.researchgate.net/publication/303551320_Stochastic_Matrix-Function_Estimators_Scalable_Big-Data_Kernels_with_High_Performance

Dr. Martin Müller Lennert

Dr. Martin Müller-Lennert is a Senior Data Scientist at InCube. He has implemented machine learning solutions in the financial industry and worked on the development of InCube's wealth tech platform. He holds both a PhD and a master’s degree in mathematics from ETH Zurich.

Milica Petrović

Milica Petrović is a Senior Data Scientist at InCube. She has professional experience applying data modelling and AI in the financial industry, working on projects from design to implementation. She holds an MSc degree in Statistics from the ETH.

Automated Data Quality Assurance with Machine Learning

Companies store massive amounts of data to derive business value from it. However, data quality issues limit the usefulness of the data. Typical issues include missing values, wrong formats, or incorrect values. The usual approach to overcome them is to implement hard-coded rules on the database entries to ensure their correctness. However, this approach is not universally applicable and does not scale.

Machine learning provides ways to significantly improve and speed up error detection, without the need to explicitly specify hard-coded rules. The secret is autoencoders – neural networks that model and reconstruct their own input. When adequately trained, they only reconstruct clean data, which exposes corrupted entries.  

Automation allows for an extension from error detection to error remediation. Using Robotic Process Automation combined with AI, also known as Intelligent Process Automation (IPA), companies can correct some types of identified errors without human involvement. To achieve this goal, IPA systems leverage not only structured data, but also unstructured data such as scanned documents. 

We apply autoencoders for detecting errors on production data and are developing IPA systems for correcting them. Alongside the specific algorithms for different data types and our findings, we will present a small demo application.

A Brief Biography

A Brief Biography

Karin Lange is Head of Corporate Foresight Management – and thus part of the Innovation department at Swiss Mobiliar. Responsible for Trend Scouting, Trend Monitoring and prospective scenario building based on the trend findings, Karin works with interdisciplinary teams within the company to explore risks and opportunities of new business fields for her company.

Prior to this position she was part of the Corporate Social Responsibility team at Swiss Mobiliar, focusing strongly on the question of Corporate Digital Responsibility. Karin accumulates 20+ years of experience in various positions, departments, industries and countries. She graduated with a Master in Political Sciences and Linguistics from the University of Konstanz.

Since 2016 Karin is the Industrial leader for the Expert Group “Data Ethics” within the Swiss Alliance for Data Intensive Services. Data Ethics and Data Privacy have been named among the top 10 trends for 2019 by Gartner and Karin explores and develops these issues both at Swiss Mobiliar and the Swiss Alliance for Data Intensive Services. Big potential and high risk at the same time, the social implications of big data and artificial intelligence have to be thoroughly examined.

 Together with different Swiss universities and companies, the Expert Group “Data Ethics” has developed an “Ethical Code of Conduct for Data-Based Value Creation” which will be presented at the conference.

Digital Responsibility: It’s all about people, not machines

A Brief Biography

Natalie Pompe is currently finishing her PhD at the University of Zurich in which she analysed the consequences of algorithmic information distribution on the democratic discourse in Switzerland. She has been working at the Berkman Klein Center at Havard Law School on research projects concerning the legal and ethical questions in AI innovation as well as global governance questions. Among other activities, she is also the legal advisor of an emerging Swiss AI start-up and she is part of the We-Publish NGO that creates a decentralised publishing infrastructure for journalists and a new media-currency.

AI Innovation in emotional analysis

AI is disrupting different industries and will change the way we live. The newest self-learning technologies promise to identify emotional states and make predictions about human intentions. Companies such as “emotion-intelligence” evaluate unconscious behaviour through the web and are used for commercial purposes. Voice analysis and language analytics are emerging in order to better classify and understand human behaviour on the emotional level. Legally speaking, this innovation triggers various questions. The concerns that scholars had with Big Data analytics and new profiling techniques situated on the intersection of personal data protection and autonomy are now elevated on a more sensitive level. – More so, the AI innovation on the level of the human psyche triggers ethical as well as sociological questions. In this short talk Natalie Pompe will outline a few legal challenges and address the pitfalls of traditional legal instruments in responding to emerging AI innovations. Furthermore, she will continue to apply the international ethical guidelines in AI innovation to a few examples from the start-up industry to illustrate the legal and ethical challenges.

ZHAW Zurich University of Applied Sciences, Technikumstrasse 71, 8401 Winterthur, Switzerland, Eulachpassage (TN, TS)

The conference venue is situated at a walking distance from the station.