Stefano Sarao Mannelli

Research Directions

• Bias generation and amplification in ML systems. What is behind biases in an ML system? What is the impact of our design choices?
• Learning differences in biological and artificial neural networks. Curriculum learning, continual learning, transfer learning, same concept but completely different results in animals and machines, why?
• Optimisation in rough landscapes. Connecting dynamics and landscape properties in optimisation.

Keen to know more? Check out the research page or you can read my most recent publications.

Latest News

You can find all the news here.

Recent publications

• A meta-learning framework for rationalizing cognitive fatigue in neural systems

  Yujun Li, Rodrigo Carrasco-Davis, Younes Strittmatter, Stefano Sarao Mannelli, Sebastian Musslick

  CogSci 2024 (Oral)

  

  

  The ability to exert cognitive control is central to human brain function, facilitating goal-directed task performance. However, humans exhibit limitations in the duration over which they can exert cognitive control -a phenomenon referred to as cognitive fatigue. This study explores a computational rationale for cognitive fatigue in continual learning scenarios: cognitive fatigue serves to limit the extended performance of one task to avoid the forgetting of previously learned tasks. Our study employs a meta-learning framework, wherein cognitive control is optimally allocated to balance immediate task performance with forgetting of other tasks. We demonstrate that this model replicates common patterns of... [Read Article]

• Bias in Motion: Theoretical Insights into the Dynamics of Bias in SGD Training

  Anchit Jain, Rozhin Nobahari, Aristide Baratin, Stefano Sarao Mannelli

  

  

  Machine learning systems often acquire biases by leveraging undesired features in the data, impacting accuracy variably across different sub-populations. Current understanding of bias formation mostly focuses on the initial and final stages of learning, leaving a gap in knowledge regarding the transient dynamics. To address this gap, this paper explores the evolution of bias in a teacher-student setup modeling different data sub-populations with a Gaussian-mixture model. We provide an analytical description of the stochastic gradient descent dynamics of a linear classifier in this setting, which we prove to be exact in high dimension. Notably, our analysis reveals how different properties... [Read Article]

• Tilting the Odds at the Lottery: the Interplay of Overparameterisation and Curricula in Neural Networks

  Stefano Sarao Mannelli, Yaraslau Ivashinka, Andrew Saxe, Luca Saglietti

  Accepted to ICML 2024

  

  

  A wide range of empirical and theoretical works have shown that overparameterisation can amplify the performance of neural networks. According to the lottery ticket hypothesis, overparameterised networks have an increased chance of containing a sub-network that is well-initialised to solve the task at hand. A more parsimonious approach, inspired by animal learning, consists in guiding the learner towards solving the task by curating the order of the examples, i.e. providing a curriculum. However, this learning strategy seems to be hardly beneficial in deep learning applications. In this work, we propose an analytical study that connects curriculum learning and overparameterisation. In... [Read Article]

• Why Do Animals Need Shaping? A Theory of Task Composition and Curriculum Learning

  Jin Hwa Lee, Stefano Sarao Mannelli, Andrew Saxe

  Accepted to ICML 2024

  

  

  Diverse studies in systems neuroscience begin with extended periods of training known as 'shaping' procedures. These involve progressively studying component parts of more complex tasks, and can make the difference between learning a task quickly, slowly or not at all. Despite the importance of shaping to the acquisition of complex tasks, there is as yet no theory that can help guide the design of shaping procedures, or more fundamentally, provide insight into its key role in learning. Modern deep reinforcement learning systems might implicitly learn compositional primitives within their multilayer policy networks. Inspired by these models, we propose and analyse... [Read Article]

• The RL Perceptron: Generalisation Dynamics of Policy Learning in High Dimensions

  Nishil Patel, Sebastian Lee, Stefano Sarao Mannelli, Sebastian Goldt, Andrew Saxe

  

  

  Reinforcement learning (RL) algorithms have proven transformative in a range of domains. To tackle real-world domains, these systems often use neural networks to learn policies directly from pixels or other high-dimensional sensory input. By contrast, much theory of RL has focused on discrete state spaces or worst-case analysis, and fundamental questions remain about the dynamics of policy learning in high-dimensional settings. Here, we propose a solvable high-dimensional model of RL that can capture a variety of learning protocols, and derive its typical dynamics as a set of closed-form ordinary differential equations (ODEs). We derive optimal schedules for the learning rates... [Read Article]

Explore other publications here.