Finnish Artificial Intelligence Society (FAIS) has given the Finnish AI Dissertation Award 2021 for Doctor of Science Vikas Verma, for his dissertation “Algorithms for Data-Efficient Training of Deep Neural Networks” completed at the Department of Computer Science, School of Science, Aalto University. The award and a 1,000€ monetary prize sponsored by Unity Technologies was granted for the winner at the Annual Symposium for Computer Science in Tampere on June 20th, 2022.
The dissertation of Verma belongs to the area of machine learning and considers deep neural networks that have achieved human-level or even super-human performance in many tasks such as computer vision, natural language processing and speech recognition. The success of these methods can be attributed to their ability to learn complex representations from raw data, assuming access to large amount of labeled data for learning the models. However, in many applications e.g. in healthcare collecting large data sets is either costly or impossible. The dissertation of Verma addresses this fundamental challenge in machine learning, developing better methods for data efficient training of deep neural networks. It introduces concepts and practical methods that provide inductive bias on how the models should behave in-between the training samples and demonstrates that the new approaches outperform conventional learning methods across several different learning settings, including semi-supervised, adversarial, and graph-based training.
The dissertation is of extremely high quality by international standards and makes significant contributions for the field. The results of the work have been presented in the highly selective leading machine learning venues, including the highest ranked conferences of Neural Information Processing Systems, International Conference on Machine Learning, International Conference on Artificial Intelligence, and International Conference on Learning Representations. These articles have already received significant international attention, best demonstrated by three of the articles already being cited for more than 250 times each.
Finnish Artificial Intelligence Society (FAIS) represents the field of AI as the official member society of the European Association for Artificial Intelligence (EurAI). It supports and promotes the advancement of public knowledge, education, research and applications of AI in Finland. It grants the Finnish AI Dissertation Award annually for an outstanding thesis in the field of AI defended in a Finnish university.
The selection of the winner was made by selection committee nominated by the board of the society, consisting of Associate Professor Arto Klami (University of Helsinki; chair), Professor Guoying Zhao (University of Oulu), and Associate Professor Ilkka Pölönen (University of Jyväskylä).
The dissertation of Vikas Verma is available at https://aaltodoc.aalto.fi/handle/123456789/59150
For more information: Chair of the Selection Committee and FAIS Board member Arto Klami, arto [.] klami [at] helsinki.fi