16.15 - 16.30 IST

12.45 - 13.00 EET

Introduction to RBC-DSAI

Opening Remarks by Prof. Balaraman Ravindran, IIT Madras

16.30 - 17.15 IST

13.00 - 13.45 EET

XAI for the Sciences: towards understanding

Keynote Talk by Prof. Klaus-Robert Müller, Technische Universität Berlin

In recent years, machine learning (ML) and artificial intelligence (AI) methods have begun to play a more and more enabling role in the sciences and in industry. In particular, the advent of large and/or complex data corpora has given rise to new technological challenges and possibilities. In his talk, Müller will first introduce explainable AI techniques and touch upon the topic of ML applications in the sciences. He will also discuss possibilities for extracting information from machine learning models to further our understanding by explaining nonlinear ML models. E.g. Machine Learning Models for Quantum Chemistry can, by applying XAI, contribute to furthering chemical understanding. Finally, Müller will briefly outline perspectives and limitations.

17.20 - 17.50 IST

13.50 - 14.20 EET

Legal and Ethical Considerations with AI-Generated Innovations

Invited Talk by Prof. Rosa Ballardini, University of Lapland

Artificial intelligence (AI) can create a vast number of physical and intangible systems already today. Increasingly, tasks and jobs traditionally done by natural persons in various fields will be fulfilled by these non-human entities. These automated or semi-automated activities have vast financial and legal consequences, pointing to the fact that the less meaningful human control, the greater the need for regulating AI entities. Besides efficiency, these developments raise several fundamental questions pertaining to one of the most important legal fields that regulate these types of innovations, namely intellectual property law (IP, IPR). To what extend is it allowed, acceptable or even desirable that AI-generated innovations are protected by some form of IPR? What about IPR infringements caused partly or even entirely by AI agents? What kind of legal and ethical considerations are to be taken into account in this assessment? By placing the study in the context of AI-generated innovations and IPR, the presentation investigates issues related to IP protection and infringement, as well as the need for regulating (or not) AI actors as legal subjects, including the feasibility for having AI as legal persons for IPR purposes.

17.55 - 18.25 IST

14.25 - 14.55 EET

Explicitising the implicit interpretability of DNNs

Invited Talk by Prof. Chandrashekar Lakshminarayanan, IIT Madras

Deep neural networks (DNNs) are still largely considered to be black boxes. Recent past has seen two kinds of efforts (i) to 'explain' the decisions of DNNs via simpler models in a post-hoc manner and (ii) to make the DNNs interpretable by design. In this backdrop, we look at DNNs with rectified linear units (ReLUs) through the lens of the dual view which we developed in our recent work. The dual view provides an analytical/empirical framework to understand the gates and the weights separately.

Using the dual view:
- we show most important information is learnt in the gates (i.e., on/off states),
- we show that it is unnecessary to interpret the computations in the weights
- we argue that the main function of the weights is in dual lifting, whereby the information in the layers gets lifted to the space of paths.

We explicitise the role of dual lifting by considering a new class of model called 'Deep Linearly Gated Networks', wherein, the pre-activations (to gates) are generated without using any non-linear activations, and then the weights are used to dual lift pre-activations. We show via experiments (on CIFAR-10, CIFAR-100) that DLGN counterparts of state-of-the-art DNNs perform almost as well as the DNNs.

We conclude by saying that
(i) dual lifting is implicit in standard DNNs with ReLUs, and seen this way one can attribute the success of standard DNNs to dual lifting.
(ii) DLGNs are more readily amenable to interpretation/explanation and merit the attention of the larger deep learning community

18.30 - 19.00 IST

15.00 - 15.30 EET

Evaluating Model Explanations

Invited Talk by Dr. Danish Pruthi, Carnegie Mellon University

While deep learning models have become increasingly accurate over the last decade, concerns about their (lack of) interpretability have taken a center stage. In response, a growing subfield on interpretability and analysis of these models has emerged. While hundreds of techniques have been proposed to “explain” predictions of models, what aims these explanations serve and how they ought to be evaluated are often unstated. In this talk, I will present a proposal to quantify the value of explanations, which draws upon argumentative theories of human reasoning that posit that (effective) explanations communicate how the decisions are made, and can help people predict how later decisions will be made.

19.05 - 19.10 IST

15.35 - 15.40 EET

Closing Remarks

16.30 - 16.45 IST

13.00 - 13.15 EET

Finnish policy level approaches for AI collaboration with India

Opening Talk by Kit Srinivasan, Aalto University and Siddharth Naithani, Business Finland

Kit Srinivasan will give an introduction to FICORE (Finnish Indian Consortia for Research and Education) network involving 23 IITs and 15 Finnish higher educational institutions.

Siddharth Naithani will address the hands-on approach from Business Finland to AI collaborations in India.

16.50 - 17.35 IST

13.20 - 14.05 EET

New Frontiers in AI and ML

Keynote Talk by Prof. Vikas Garg, Aalto University

Artificial Intelligence (AI) and Machine Learning (ML), including deep learning (DL), continue to transform the world around us. However, in order to avoid the risk of "plateauing out", we need to make the next leap. In this talk, I will provide an overview of some exciting work we've been doing to design cutting-edge technologies and push the boundaries of what can be achieved with AI/ML/DL in several important domains such as biopharma, IoT, quantum computing, multiagent systems, and e-commerce. I'll also mention relevant research being conducted at FCAI.

For the benefit of our audience - both within academia and industry - I'll present several motivating examples from recent AI/ML/DL success stories, and sketch how critically dissecting the (complex) problems of interest for scientists and industrial practitioners, and combining intuition with mathematical rigor is often the key to making these scientific advances.

17.40 - 18.10 IST

14.10 - 14.40 EET

Accelerating drug design with AI

Invited Talk by Prof. Ola Engkvist, AstraZeneca Gothenburg

Artificial Intelligence has become impactful during the last few years in chemistry and the life sciences, pushing the scientific boundaries forward as exemplified by the recent success of AlphaFold2. In this presentation, I will provide an overview of how AI has impacted drug design in the last few years, where we are now and what progress we can reasonably expect in the coming years. The presentation will have a focus on deep learning based molecular de novo design, however, also aspects of synthesis prediction, molecular property predictions and chemistry automation will be covered.

18.15 - 18.45 IST

14.45 - 15.15 EET

Guiding Medicine Design with AI

Invited Talk by Julius Sipilä, Orion Pharma R&D

An overview will be presented how AI/ML methods are taken into use and developed for Medicine Design in Orion Pharma R&D for different tasks: molecular property predictions, automated ML, AI-assisted structure-based virtual screening and generative models for designing novel small molecule drugs.

18.50 - 19.20 IST

15.20 - 15.50 EET

Learning to Predict Drug Combination Responses

Invited Talk by Prof. Juho Rousu, Aalto University

Combinatorial treatments involving two or more drugs have become a standard of care for various complex diseases, including tuberculosis, malaria, HIV and other viral infections, as well as most of the advanced cancers. High-throughput screening in preclinical model systems (e.g. cancer cell lines or viral infection models) is the state-of-the-art approach to systematically identify candidate drug combinations. However, due to the exponential number of possible drug combinations and the extensive heterogeneity of the target systems, computational methods, and in particular, machine learning, are critically needed to guide the discovery of effective combinations to be prioritized for further pre-clinical validation and clinical development.

In this talk, I present our recent work on a highly accurate machine learning framework for predicting the responses of drug combinations in preclinical studies. Our methods model the cell context-specific drug interactions through higher-order tensors, and efficiently learns latent factors of the tensor using powerful factorization machines and latent tensor reconstruction. The approach enables our methods to leverage information from previous experiments performed on similar drugs and cells when predicting responses of new combinations in so far untested cells; thereby, it achieves highly accurate predictions despite sparsely populated data tensors. In conclusion, the presentation discusses the open challenges and potential extensions of the methods.

19.25 - 19.30 IST

15.55 - 16.00 EET

Closing Remarks

16.30 - 16.45 IST

13.00 - 13.15 EET

Engage and collaborate with FCAI

Opening Talk by Terhi Kajaste, Finnish Center for Artificial Intelligence

Finnish Center for Artificial Intelligence (FCAI) is a community of experts that worldwide brings together top talents in academia, industry and the public sector to solve real-life problems using both existing and novel AI. Our mission is to create AI that is data-efficient, trustworthy and ethically sustainable, and that can operate with people in an understandable manner. Our activities span a wide spectrum of fields. We collaborate with academic, industrial and societal actors, and have high-volume strategic initiatives as well as numerous smaller-scale bottom-up projects. We also have a wide and outstandingly popular education program. In her talk, Kajaste will invite the audience into collaboration with FCAI and give concrete case examples of various kinds of successful collaboration

16.50 - 17.35 IST

13.20 - 14.05 EET

Empowering Intelligence to the Edge of Network

Keynote Talk by Prof. Sasu Tarkoma, University of Helsinki

17.40 - 18.10 IST

14.10 - 14.40 EET

Data access and management for distributed multi-X edge application

Invited Talk by Kimmo Hätönen, Nokia Bell Labs, Finland

In a multi-x edge environment, it is a challenge to timely share partially overlapping data streams for multiple AI functions in several edges doing simultaneous inference on the data. The challenge becomes even harder, if the same data is needed for off-line training done later elsewhere. In this presentation, I will discuss some aspects of the problem and present related findings of recent research projects that studied domains of air quality monitoring and energy ecosystems.

18.15 - 18.45 IST

14.45 - 15.15 EET

Accelerating AI algorithms on the edge: from software to hardware challenges

Invited Talk by Prof. Martin Andraud, Aalto University

18.50 - 19.20 IST

15.20 - 15.50 EET

Leveraging Machine Learning for Spectrum Sharing in Wireless Networks

Invited Talk by Prof. Suzan Bayhan, University of Twente

19.25 - 19.30 IST

15.55 - 16:00 EET

Closing Remarks

16.30 - 16.45 IST

13.00 - 13.15 EET

Spotlight Talk 1

by Shivam Gupta, Ganesh Ghalme, Narayanan C. Krishnan, Shweta Jain

Title: Efficient Algorithms For Fair Clustering with a New Fairness Notion

16.50 - 17.05 IST

13.20 - 13.35 EET

Spotlight Talk 2

by Calvin Guillot, Suyog Chandramouli, Antti Oulasvirta

Title: Human-in-the-loop Hyperparameter Tuning of Deep Nets to Improve Explainability of Classifications

17.10 - 17.25 IST

13.40 - 13.55 EET

Spotlight Talk 3

by Siddhesh Nesarikar, Omkar Nath, Ashwin K Raghu, Gokul S Krishnan, Mounisai Siddhartha Middela, Gitakrishnan Ramadurai

Title: Travel Time Estimation for Emergency Medical Service Vehicles

17.30 - 17.45 IST

14.00 - 14.15 EET

Spotlight Talk 4

by Adithya Ramesh, Balaraman Ravindran

Title: Hamiltonian Model Based Reinforcement Learning for Robotics

17.50 - 17.55 IST

14.20 - 14.25 EET

Taylor and Francis presentation

17.55 - 18.00 IST

14.25 - 14.30 EET

Best Spotlight Talk Award

from Taylor and Francis

18.05 - 19.30 IST

14.35 - 16.00 EET

Gather poster sessions & mentoring session

  • Arun Rajkumar, Assistant Professor, IIT Madras
  • Ashwin Rao, Assistant Professor, University of Helsinki
  • Vikas Garg, Assistant Professor, Aalto University
  • Harish Guruprasad, Assistant Professor, IIT Madras
  • Chandrashekar Lakshminarayanan, Assistant Professor, IIT Madras
  • Amauri Holanda De Souza Junior, Postdoctoral Researcher, Aalto University
  • Naser Motlaugh, Postdoctoral Researcher, University of Helsinki
  • Pranvera Kortoçi, Postdoctoral Researcher, University of Helsinki