Andy Radler

Proud of our Geometry-Informed Neural Net (GINN🍸) which generates diverse shapes without training data. Paper: arxiv.org/abs/2402.14009 Great work and thanks to my amazing team mates and supervisors @artuursberzins @e_volkmann @SebSanokowski @HochreiterSepp @jo_brandstetter

👉 GINNs: A new method to train shape-generative models using design constraints, producing diverse geometries with a unique diversity constraint. 👉GenTO: Applies GINNs to topology optimization, using a solver-in-the-loop to create diverse and structurally compliant shapes.

I had the privilege of presenting my latest work at the renowned CRUNCH seminar at Brown University. 👉 Geometry-informed Neural Networks (GINNs) 👉 Generative Topology Optimization (GenTO) Video: youtu.be/bTsbZcAlfXY?t=… With @e_volkmann @artuursberzins @jo_brandstetter

Geometry-informed Neural Networks: arxiv.org/abs/2402.14009 Generative Topology Optimization: arxiv.org/abs/2502.13174

Super hyped to share NeuralDEM -- the first real-time simulation of industrial particulate flows. NeuralDEM replaces Discrete Element Method (DEM) routines and coupled (CFD-DEM) multiphysics simulations. 🧵 📜: arxiv.org/abs/2411.09678 🖥️: nx-ai.github.io/NeuralDEM/

Thanks to all the folks visiting our poster on Geometry-informed Neural Networks (GINNs) at the workshop on Geometry-grounded Representation Learning and Generative Modelling. With @artuursberzins @SebSanokowski @jo_brandstetter @HochreiterSepp #ICML2024 #GenerativeAI

Thanks for the fantastic collaboration @artuursberzins @SebSanokowski @jo_brandstetter @HochreiterSepp

We introduce Geometry-Informed Neural Networks to train shape generative models without any data (!!), combining learning under constraints, neural fields as a suitable representation, and generating diverse solutions to under-determined problems: 🖥️: arturs-berzins.github.io/GINN/

SymbolicAI A framework for logic-based approaches combining generative models and solvers paper page: huggingface.co/papers/2402.00… introduce SymbolicAI, a versatile and modular framework employing a logic-based approach to concept learning and flow management in generative processes. SymbolicAI enables the seamless integration of generative models with a diverse range of solvers by treating large language models (LLMs) as semantic parsers that execute tasks based on both natural and formal language instructions, thus bridging the gap between symbolic reasoning and generative AI. We leverage probabilistic programming principles to tackle complex tasks, and utilize differentiable and classical programming paradigms with their respective strengths. The framework introduces a set of polymorphic, compositional, and self-referential operations for data stream manipulation, aligning LLM outputs with user objectives. As a result, we can transition between the capabilities of various foundation models endowed with zero- and few-shot learning capabilities and specialized, fine-tuned models or solvers proficient in addressing specific problems. In turn, the framework facilitates the creation and evaluation of explainable computational graphs. We conclude by introducing a quality measure and its empirical score for evaluating these computational graphs, and propose a benchmark that compares various state-of-the-art LLMs across a set of complex workflows. We refer to the empirical score as the "Vector Embedding for Relational Trajectory Evaluation through Cross-similarity", or VERTEX score for short. The framework codebase and benchmark are linked below.

Personal update: last month, I re-joined the group of my mentor @HochreiterSepp and my amazing colleague @gklambauer in Linz, opening my own group "AI for data-driven simulations". We all share the vision to create a large-scale AI ecosystem in Linz. Big news to come soon 🚀

One of the clearest signs of learning is rethinking your assumptions and revising your opinions. 21 things I rethought in 2021: a thread...

In our recent work, we address the problem of parameter choice in unsupervised domain adaptation by aggregation. Paper (selected for oral presentation at #ICLR2023): openreview.net/forum?id=M95oD… [1/2]

A Dataset Perspective on Offline Reinforcement Learning abs: arxiv.org/abs/2111.04714 github: github.com/ml-jku/Offline…

ArXiv arxiv.org/abs/2204.07118: New simple data-augmentation for Vision Transformers (ViTs): Grayscale, Solarization, Gaussian Blur. Suggests simple random crops. Outperforms by a large margin previous fully supervised training recipes for ViTs.