Author name cluster

Giorgio Giannone

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

6 papers

2 author rows

TMLR Journal 2025 Journal Article

NITO: Neural Implicit Fields for Resolution-free and Domain-Adaptable Topology Optimization

Amin Heyrani Nobari
Lyle Regenwetter
Giorgio Giannone
Faez Ahmed

Structural topology optimization plays a crucial role in engineering by determining the optimal material layout within a design space to maximize performance under given constraints. We introduce Neural Implicit Topology Optimization (NITO), a deep learning regression approach to accelerate topology optimization tasks. We demonstrate that, compared to state-of-the-art diffusion models, NITO generates structures that are under 15% as structurally sub-optimal and does so ten times faster. Furthermore, we show that NITO is entirely resolution-free and domain-agnostic, offering a more scalable solution than the current fixed-resolution and domain-specific diffusion models. To achieve this state-of-the-art performance, NITO combines three key innovations. First, we introduce the Boundary Point Order-Invariant MLP (BPOM), which represents loads and supports in a sparse and domain-agnostic manner, allowing NITO to train on variable conditioning, domain shapes, and mesh resolutions. Second, we adopt a neural implicit field representation, which allows NITO to synthesize topologies of any shape or resolution. Finally, we propose an inference-time refinement step using a few steps of gradient-based optimization to enable NITO to achieve results comparable to direct optimization methods. These three innovations empower NITO with a precision and versatility that is currently unparalleled among competing deep learning approaches for topology optimization. Code & Data: https://github.com/ahnobari/NITO_Public

PDF Details

TMLR Journal 2024 Journal Article

Constraining Generative Models for Engineering Design with Negative Data

Lyle Regenwetter
Giorgio Giannone
Akash Srivastava
Dan Gutfreund
Faez Ahmed

Generative models have recently achieved remarkable success and widespread adoption in society, yet they still often struggle to generate realistic and accurate outputs. This challenge extends beyond language and vision into fields like engineering design, where safety-critical engineering standards and non-negotiable physical laws tightly constrain what outputs are considered acceptable. In this work, we introduce two approaches to guide models toward constraint-satisfying outputs using `negative data' -- examples of what to avoid. Our negative data generative models (NDGMs) outperform state-of-the-art NDGMs by 4x in constraint satisfaction and easily outperform classic generative models using 8x less data in certain problems. To demonstrate this, we rigorously benchmark our NDGMs against 14 baseline models across numerous synthetic and real engineering problems, such as ship hulls with hydrodynamic constraints and vehicle design with impact safety constraints. Our benchmarks showcase both the best-in-class performance of our new NDGM models and the widespread dominance of NDGMs over classic generative models in general. In doing so, we advocate for the more widespread use of NDGMs in engineering design tasks.

PDF Details

NeurIPS Conference 2024 Conference Paper

Reparameterized Multi-Resolution Convolutions for Long Sequence Modelling

Harry J. Cunningham
Giorgio Giannone
Mingtian Zhang
Marc P. Deisenroth

Global convolutions have shown increasing promise as powerful general-purpose sequence models. However, training long convolutions is challenging, and kernel parameterizations must be able to learn long-range dependencies without overfitting. This work introduces reparameterized multi-resolution convolutions ($\texttt{MRConv}$), a novel approach to parameterizing global convolutional kernels for long-sequence modeling. By leveraging multi-resolution convolutions, incorporating structural reparameterization and introducing learnable kernel decay, $\texttt{MRConv}$ learns expressive long-range kernels that perform well across various data modalities. Our experiments demonstrate state-of-the-art performance on the Long Range Arena, Sequential CIFAR, and Speech Commands tasks among convolution models and linear-time transformers. Moreover, we report improved performance on ImageNet classification by replacing 2D convolutions with 1D $\texttt{MRConv}$ layers.

PDF Details DOI

NeurIPS Conference 2023 Conference Paper

Aligning Optimization Trajectories with Diffusion Models for Constrained Design Generation

Giorgio Giannone
Akash Srivastava
Ole Winther
Faez Ahmed

Generative models have significantly influenced both vision and language domains, ushering in innovative multimodal applications. Although these achievements have motivated exploration in scientific and engineering fields, challenges emerge, particularly in constrained settings with limited data where precision is crucial. Traditional engineering optimization methods rooted in physics often surpass generative models in these contexts. To address these challenges, we introduce Diffusion Optimization Models (DOM) and Trajectory Alignment (TA), a learning framework that demonstrates the efficacy of aligning the sampling trajectory of diffusion models with the trajectory derived from physics-based iterative optimization methods. This alignment ensures that the sampling process remains grounded in the underlying physical principles. This alignment eliminates the need for costly preprocessing, external surrogate models, or extra labeled data, generating feasible and high-performance designs efficiently. We apply our framework to structural topology optimization, a fundamental problem in mechanical design, evaluating its performance on in- and out-of-distribution configurations. Our results demonstrate that TA outperforms state-of-the-art deep generative models on in-distribution configurations and halves the inference computational cost. When coupled with a few steps of optimization, it also improves manufacturability for out-of-distribution conditions. DOM's efficiency and performance improvements significantly expedite design processes and steer them toward optimal and manufacturable outcomes, highlighting the potential of generative models in data-driven design.

PDF Details

ICML Conference 2023 Conference Paper

Unifying Molecular and Textual Representations via Multi-task Language Modelling

Dimitrios Christofidellis
Giorgio Giannone
Jannis Born
Ole Winther
Teodoro Laino
Matteo Manica

The recent advances in neural language models have also been successfully applied to the field of chemistry, offering generative solutions for classical problems in molecular design and synthesis planning. These new methods have the potential to fuel a new era of data-driven automation in scientific discovery. However, specialized models are still typically required for each task, leading to the need for problem-specific fine-tuning and neglecting task interrelations. The main obstacle in this field is the lack of a unified representation between natural language and chemical representations, complicating and limiting human-machine interaction. Here, we propose the first multi-domain, multi-task language model that can solve a wide range of tasks in both the chemical and natural language domains. Our model can handle chemical and natural language concurrently, without requiring expensive pre-training on single domains or task-specific models. Interestingly, sharing weights across domains remarkably improves our model when benchmarked against state-of-the-art baselines on single-domain and cross-domain tasks. In particular, sharing information across domains and tasks gives rise to large improvements in cross-domain tasks, the magnitude of which increase with scale, as measured by more than a dozen of relevant metrics. Our work suggests that such models can robustly and efficiently accelerate discovery in physical sciences by superseding problem-specific fine-tuning and enhancing human-model interactions.

Details

ICML Conference 2022 Conference Paper

SCHA-VAE: Hierarchical Context Aggregation for Few-Shot Generation

Giorgio Giannone
Ole Winther

A few-shot generative model should be able to generate data from a novel distribution by only observing a limited set of examples. In few-shot learning the model is trained on data from many sets from distributions sharing some underlying properties such as sets of characters from different alphabets or objects from different categories. We extend current latent variable models for sets to a fully hierarchical approach with an attention-based point to set-level aggregation and call our method SCHA-VAE for Set-Context-Hierarchical-Aggregation Variational Autoencoder. We explore likelihood-based model comparison, iterative data sampling, and adaptation-free out-of-distribution generalization. Our results show that the hierarchical formulation better captures the intrinsic variability within the sets in the small data regime. This work generalizes deep latent variable approaches to few-shot learning, taking a step toward large-scale few-shot generation with a formulation that readily works with current state-of-the-art deep generative models.

Details