NORTHAMPTON, MA / ACCESSWIRE / July 9, 2024 / Qualcomm:
Each winner receives mentoring and $40,000 in research funding
Qualcomm Technologies, Inc. announced the recipients of the Qualcomm Innovation Fellowship (QIF) Europe program, now in its 15th year: Dongqing Wang (EPFL), Neelu S. Kalani (EPFL), Chaitanya K. Joshi (University of Cambridge), Runa Eschenhagen (University of Cambridge) and Afra Amini (ETH Zurich)
The QIF is an annual program that aims to recognize, reward and mentor the most innovative engineering PhD students from Europe, India and the United States. The European program rewards young researchers of excellence in the fields of artificial intelligence and cybersecurity with individual prizes of $40,000 and dedicated mentors from the Qualcomm Technologies research team.
“This year, we received nearly 50% more applications than last year, which is a testament to the growth of the machine learning field, as well as the need to further secure our rapidly developing software and hardware,” said Michael Hofmann, Senior Director of Engineering at Qualcomm Technologies Netherlands BV. “This year’s proposals ranged from foundational algorithms and large language models to exciting applications such as extended reality, generalizable computer vision, RNA design, and more. We are honored to mentor all of the winners in their research.”
The seventeen finalists were doctoral students from ETH Zurich, Imperial College London, University of Edinburgh, University of Tübingen, University of Cambridge, University of Oxford, CISPA Helmholtz Center for Information Security, TU Delft and the Czech Technical University.
After careful consideration, the following five winners were selected for their exceptional proposals:
“Towards a visually plausible and controllable 360° virtual reality” – Dongqing Wang
Creating visually plausible and controllable virtual reality (VR) of the real world is essential to enable quality immersive experiences in extended reality (XR) applications. NeuralRadiance Fields (NeRFs) and their variants can model 360-degree real-world scenes for photorealistic view synthesis with low memory storage. Therefore, they have the potential to become widely accessible 3D world representations. However, the implicit nature of their underlying representation makes it difficult to directly edit a 3D NeRF scene. For controllability, we propose a 3-component strategy to enable an editing system on NeRF. The outcome of this proposal will aim to enable visually plausible and controllable 360-degree VR to enhance interaction with the virtual world.
“FlashPoint: A Dynamic and Secure Root of Trust” – Neelu S. Kalani
The field of confidential computing has seen significant advances over the past decade. Even as confidential computing evolves, minimizing the trusted computing base (TCB) remains crucial. So far, little attention has been paid to removing platform-specific firmware, which runs with the highest privileges alongside security monitors that provide confidential computing guarantees, from the TCB. In the meantime, many vulnerabilities in large and buggy platform-specific firmware have been exploited (e.g. to disclose platform secret keys) to compromise the security of the entire system. We propose FlashPoint, a dynamic root-of-trust solution for RISC platforms. It includes ISA extensions to enable secure transitions between trusted and untrusted code running in the highest privilege mode, without introducing a new privilege mode.
“Geometric Generative Models for 3D RNA Design” – Chaitanya K. Joshi
This proposal aims to develop the first deep learning framework for 3D RNA design. I will describe an execution plan for an RNA-centric 3D generative model that builds on best practices that have revolutionized protein design. I will explain why AlphaFold is not enough, how to address RNA-specific modeling challenges, and why integrating the inductive biases that determine RNA structure is essential to develop tailored generative models for RNA design.
“A journey towards controlled and efficient text generation” – Afra Amini
Imagine a scenario in which you use a language model of your choice to generate a fictional story. You ask the language model to generate a story about a TikZ unicorn that comes to life, and the model generates a story. While fascinated by this amazing technology, you realize that the story is too short for your purpose, the language used is very formal and not suitable for your target audience, and the sentence structures are too complex. How can you systematically control these aspects in the generated story? Which knobs in the model should you tweak to ensure that the generations satisfy the desired constraints? In this work, we explore recent advances in two research directions to systematically control text generation. We also demonstrate how these two approaches can be unified as different methods to approximate the same goal.
“Towards an understanding of curvature matrices in deep learning” – Runa Eschenhagen
Many algorithms that attempt to address the shortcomings of deep learning rely on approximations of the Hessian of the loss with respect to the neural network parameters or other related quantities, called curvature matrices. This includes second-order optimization methods to improve training efficiency, influence functions for data attribution, pruning and compression methods, quantification of predictive uncertainty, etc. However, the effect of curvature approximation on the performance of downstream tasks is not well understood. My proposal aims to improve our understanding of commonly used curvature approximations, in particular variants of K-FAC. This has the potential to directly impact all applications that rely on these approximations, inform the design of new approximations, provide insights into training dynamics, and lay the foundation for new theoretical explanations.
About Qualcomm
Qualcomm relentlessly innovates to deliver intelligent computing everywhere, helping the world solve some of its biggest challenges. Our proven solutions drive transformation across major industries, and our Snapdragon® branded platforms deliver extraordinary customer experiences. With nearly 40 years of leadership in setting industry standards and creating breakthrough technologies, we deliver cutting-edge AI, low-power, high-performance computing, and unmatched connectivity. Together with our ecosystem partners, we enable next-generation digital transformation to enrich lives, improve businesses, and advance societies. At Qualcomm, we are engineering human progress.
Qualcomm Incorporated includes our licensing business, QTL, and the vast majority of our patent portfolio. Qualcomm Technologies, Inc., a subsidiary of Qualcomm Incorporated, operates, together with its subsidiaries, substantially all of our engineering and research and development functions and substantially all of our products and services businesses, including our QCT semiconductor business. Snapdragon and Qualcomm branded products are products of Qualcomm Technologies, Inc. and/or its subsidiaries. Qualcomm patented technologies are licensed from Qualcomm Incorporated.
Contact details:
Mauricio Lopez-Hodoyan
Investor Relations
1-858-658-4813
(protected email)
Clare Conley
Media Relations
1-858-845-5959
(protected email)
Explore additional media and other Qualcomm ESG stories at 3blmedia.com.
Contact information:
Spokesperson: Qualcomm
Website: https://www.3blmedia.com/profiles/qualcomm
E-mail: (protected email)
SOURCE: Qualcomm
