Author name cluster

Francesco Bertolotti

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.

4 papers

1 author row

ECAI Conference 2025 Conference Paper

SHAC++: A Neural Network to Rule All Differentiable Simulators

Francesco Bertolotti
Gianluca Aguzzi
Walter Cazzola
Mirko Viroli

Reinforcement learning (RL) algorithms show promise in robotics and multi-agent systems but often suffer from low sample efficiency. While methods like SHAC leverage differentiable simulators to improve efficiency, they are limited to specific settings: they require fully differentiable environments, including transition and reward functions, and have primarily been demonstrated in single-agent scenarios. To overcome these limitations, we introduce SHAC++, a novel framework inspired by SHAC. SHAC++ removes the need for differentiable simulator components by using neural networks to approximate the required gradients, training these networks alongside the standard policy and value networks. This enables the core SHAC approach to be applied in both non-differentiable and multi-agent environments. We evaluate SHAC++ on challenging multi-agent tasks from the VMAS suite, comparing it against SHAC (where applicable) and PPO, a standard algorithm for non-differentiable settings. Our results demonstrate that SHAC++ significantly outperforms PPO in both single- and multi-agent scenarios. Furthermore, in differentiable environments where SHAC operates, SHAC++ achieves comparable performance despite lacking direct access to simulator gradients, thus successfully extending SHACs benefits to a broader class of problems. The full implementation is openly available at https: //github. com/f14-bertolotti/shacpp.

Details

ICML Conference 2024 Conference Paper

By Tying Embeddings You Are Assuming the Distributional Hypothesis

Francesco Bertolotti
Walter Cazzola

In this work, we analyze both theoretically and empirically the effect of tied input-output embeddings—a popular technique that reduces the model size while often improving training. Interestingly, we found that this technique is connected to Harris (1954)’s distributional hypothesis—often portrayed by the famous Firth (1957)’s quote “a word is characterized by the company it keeps”. Specifically, our findings indicate that words (or, more broadly, symbols) with similar semantics tend to be encoded in similar input embeddings, while words that appear in similar contexts are encoded in similar output embeddings (thus explaining the semantic space arising in input and output embedding of foundational language models). As a consequence of these findings, the tying of the input and output embeddings is encouraged only when the distributional hypothesis holds for the underlying data. These results also provide insight into the embeddings of foundation language models (which are known to be semantically organized). Further, we complement the theoretical findings with several experiments supporting the claims.

Details

EUMAS Conference 2020 Conference Paper

"Roads? Where We're Going We Don't Need Roads. " Using Agent-Based Modeling to Analyze the Economic Impact of Hyperloop Introduction on a Supply Chain

Francesco Bertolotti
Riccardo Occa

Abstract The opportunity to connect distant areas in quick and economical ways has always been a critical element for trade and economic development. Significant progress is often accompanied by the emergence of new transport models, as it happened in 21st century China with high-speed railways, a driving force for its economic growth. In 2013, the Hyperloop Alpha white paper publication presented the opportunity for a significant innovation in transportation, which could have an even more disruptive effect. In this paper, we will estimate the possible impact on a Supply Chain of constructing a Hyperloop line for goods transportation. We develop an agent-based model of a simple Supply Chain system to simulate the introduction of a faster transport line. We observed that a positive relationship exists between the introduction of a faster line and the performance of the firms near the cities connected by the new line. We conclude that the adoption of a Hyperloop could significantly affect the region in which it is implemented. Since the technology is in its infancy, there is still room for further research.

Details DOI

EUMAS Conference 2020 Conference Paper

Sensitivity to Initial Conditions in Agent-Based Models

Francesco Bertolotti
Angela Locoro
Luca Mari

Abstract In the last thirty years, agent-based modelling has become a well-known technique for studying and simulating dynamical systems. Still, there are some open issues to be addressed. One of these is the substantial absence of studies about the sensitivity to initial conditions, that is the effect of small variations at the beginning of simulation on the macro-level behaviour of the model. The goal of this preliminary work is to explore how a single modification on one agent affects the evolution of the simulation. Through the analysis of two deterministic models (a simple market model and Reynolds’ flocking model), we obtain two main results. First, we observe that the impact of the variation of a single initial condition on the simulation behaviour is high in both models. Second, there is evidence of an at least qualitative relation between some general agent-based model settings (numerosity of agents in the model and rate of connections between agents) and the sensitivity to the modified initial condition. We conclude that at least some significant classes of agent-based models are affected by a high sensitivity to initial conditions that have a negative effect on the predictive power of simulations.

Details DOI