Fabio Peruzzo | TransferLab — appliedAI Institute

Fabio Peruzzo

After receiving his PhD in Applied Mathematics with a specialization in complex systems, Fabio was drawn to the potential of the burgeoning field of machine learning. This interest led him to transition into AI engineering. His professional journey includes roles at several startups in the UK. As of 2021, he has been contributing his expertise to the appliedAI Transferlab team. His primary areas of research encompass data valuation, interpretable AI, and reinforcement learning.

Fabio's work

Practical anomaly detection

An two-day introduction to unsupervised ML techniques for anomaly detection and their strengths and weaknesses in different application areas.

Interpreting CLIP's Image Representation via Text-based Decomposition

Interpreting the output of neural networks is often challenging because it entails putting into words patterns that may not be easily …

Studying LLMs with Influence Functions

Fabio Peruzzo, Senior AI Engineer at appliedAI Initiative, will talk about a recent work on the use of influence functions to study the …

DataPerf, Benchmarks for Data-Centric AI Development

Much of the recent advancements in AI have been fueled by the pursuit to enhance performance on benchmarks with fixed datasets, thus …

Studying Large Language Model Generalization with Influence Functions

Influence functions are a tool to quantify the impact of each training sample on a model&rsquo;s predictions, thereby assisting in the …

Jump-Start Reinforcement Learning

Reinforcement learning often presents a challenge when training an agent from scratch, as reward sparsity and exploration issues can impede …

Safe and efficient deep reinforcement learning

This 2-day dive into deep RL for decision and control is suitable for engineers who want to solve real-world control problems with efficient …

Influence functions and Data Pruning: from theory to non-convergence

Today’s session brings Influence Functions under the spotlight - the theory, non-convergence issues, and uses for data pruning. Fabio will …

Post-Hoc Concept Bottleneck Models

Concept Bottleneck models are a set of intrinsically interpretable models that rely on high level concepts to make predictions. They often …

Beyond neural scaling laws: beating power law scaling via data pruning

Large neural networks are very &ldquo;data hungry&rdquo;, but data pruning holds the promise to alleviate this dependence. A recent paper …

What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimation

When data samples are difficult to learn, neural networks tend to memorize their labels rather than infer useful features. This has been …

Robustly representing uncertainty through sampling in deep neural networks

In the previous instalments of our seminar series we have explored how bayesian methods can be used to estimate the epistemic uncertainty of …

Planning with Diffusion for Flexible Behavior Synthesis

Diffusion models are finding their way into more machine learning domains. This paper introduces a framework for applying them to …

Prompt-to-Prompt Image Editing with Cross Attention Control

Generating images from text is complicated by the fact that small changes in prompt gives very different results. This new study shows how …

pyDVL: the python Data Valuation Library

pyDVL strives to offer reference implementations of algorithms for data valuation and influence function computation, with an emphasis on …

Neural Fields in Visual Computing and Beyond

A great entry-level review to coordinate-based neural networks (a.k.a. neural fields), a technique that has recently shown widespread …

Methods and issues in explainable AI

A two-day workshop for ML practitioners wishing to make their models better understandable for themselves and decision makers.

Brax - A Differentiable Physics Engine for Large Scale Rigid Body Simulation

The development of versatile robots demands a better understanding of physics simulators. A recent line of research focuses on engines whose …

Simulation in Robotics Opportunities Challenges Suggestions

Despite great advancements in physics based simulators for robotics over the last few years, a big sim2real gap remains in the performance …

Denoising Diffusion and Score Based Generative Models

This pill gives a short overview of the development of diffusion models over the last 7 years. It contains several references useful for …

Progressive Distillation for Fast Sampling of Diffusion Models

The generation process of diffusion models can be sped up by using a slow teacher model for training a faster student - the latter …

Our take on ICLR 2022 - Part 1

ICLR is one of the most important and prestigious conferences in the ML community. It spans all ML topics, from embeddings to Graph neural …

Score-Based Generative Modeling with Critically-Damped Langevin Diffusion

In this rather technical paper, a new dynamics is proposed for the sampling of diffusion models. Instead of sampling the de-noised image …

Energy-Based Learning for Cooperative Games, with Applications to Valuation Problems in Machine Learning

A theoretical framework connecting data valuation scores, game theory and energy based models is presented. Common criteria like Shapley or …

Resolving Training Biases via Influence-based Data Relabeling

Influence functions are used to correct corrupt labels in a dataset that would significantly decrease a trained model&rsquo;s performance. …

Analytic-DPM: an Analytic Estimate of the Optimal Reverse Variance in Diffusion Probabilistic Models

An analytic form for the de-noising process of diffusion neural networks is found in this important work. Monte Carlo sampling with the …

GeoDiff: A Geometric Diffusion Model for Molecular Conformation Generation

Diffusion models are used for predicting the stable 3d conformations of a molecule from its 2d graph.

Score-Based Generative Modeling through Stochastic Differential Equations

A seminal contribution to the field of diffusion models, here a connection between de-noising, score-matching and stochastic differential …

Influence functions are a tool to quantify the impact of each training sample on a model’s predictions, thereby assisting in the …

Large neural networks are very “data hungry”, but data pruning holds the promise to alleviate this dependence. A recent paper …

Influence functions are used to correct corrupt labels in a dataset that would significantly decrease a trained model’s performance. …