Ferran Alet

I am a PhD student at MIT CSAIL, where I work on machine learning with Leslie Kaelbling and Tomas Lozano-Perez, and Josh Tenenbaum. I am also the organizer of the MIT Embodied Intelligence Seminar.

Mentoring: I love mentoring students and working with them. I was recently honored with the MIT Outstanding Direct Mentor Award '21 (given to 2 PhDs across all MIT). If you're interested in a UROP or an MEng don't hesitate to reach out!

Currently mentoring Masters: Dylan Doblar, Shreyas Kapur (with Josh Tenenbaum). Undergrads: Javier Lopez-Contreras, Jan Olivetti.

Past mentees: Masters: Martin Schneider (moved to MIT PhD), Erica Weng (moved to CMU PhD), Adarsh K. Jeewajee (moved to Stanford PhD), Paolo Gentili. Undergrads: Max Thomsen(with Maria Bauza), Catherine Wu(with Yilun Du), Nurullah Giray Kuru, Margaret Wu, Edgar Moreno, Shengtong Zhang, Patrick John Chia , Catherine Zeng, Scott Perry.

Email  /  CV  /  Google Scholar  /  LinkedIn

profile photo
Invited talks
  • Meta-learning and multi-agent workshop 2020: Meta-learning and compositionality
  • ICML Graph Neural Network workshop 2020: Scaling from simple problems to complex problems using modularity.
  • Pierre-Yves Oudeyer's lab(INRIA) 2020: Meta-learning curiosity algorithms.
  • MIT Machine Learning Tea 2019: Meta-learning and combinatorial generalization.
  • Pieter Abbeel's lab (UC Berkeley) 2019: Meta-learning structure (slides here).
  • KR2ML@IBM Workshop 2019: Graph Element Networks (slides here, video of very similar talk at ICML).
Papers
Noether Networks: meta-learning useful conserved quantities
Ferran Alet* , Dylan Doblar*, Allan Zhou, Joshua B. Tenenbaum, Kenji Kawaguchi, Chelsea Finn (final order TBD)
NeurIPS 2021  

We propose to encode symmetries as conservation tailoring losses and meta-learn them from raw inputs in sequential prediction problems.

Tailoring: encoding inductive biases by optimizing unsupervised objctives at prediction time
Ferran Alet , Maria Bauza, Kenji Kawaguchi, Nurullah Giray Kuru, Tomás Lozano-Pérez, Leslie Pack Kaelbling,
NeurIPS 2021; Workshop version was a Spotlight at the physical inductive biases workshop

We optimize unsupervised losses for the current input. By optimizing where we act, we bypass generalization gaps and can impose a wide variety of inductive biases.

A large-scale benchmark for few-shot program induction and synthesis
Ferran Alet* , Javier Lopez-Contreras*, James Koppel, Maxwell Nye, Armando Solar-Lezama, Tomás Lozano-Pérez, Leslie Pack Kaelbling, Joshua B. Tenenbaum
ICML 2021  
website

We generate a large quantity of diverse real programs by running code instruction-by-instruction and obtain I/O pairs for 200k subprograms.

Meta-learning curiosity algorithms
Ferran Alet* , Martin Schneider*, Tomás Lozano-Pérez, Leslie Pack Kaelbling
ICLR 2020
code, press

By meta-learning programs instead of neural network weights, we can increase meta-learning generalization. We discover new algorithms in simple environments that generalize to complex ones.

Neural Relational Inference with Fast Modular Meta-learning
Ferran Alet , Erica Weng, Tomás Lozano-Pérez, Leslie Pack Kaelbling
NeurIPS, 2019  
code

We frame neural relational inference as a case of modular meta-learning and speed up the original modular meta-learning algorithms by two orders of magnitude, making them practical.

Omnipush: accurate, diverse, real-world dataset of pushing dynamics with RGB-D video
Maria Bauza, Ferran Alet Yen-Chen Lin, Tomás Lozano-Pérez, Leslie Pack Kaelbling, Phillip Isola, Alberto Rodriguez
IROS, 2019
project website / code / data / press

Diverse dataset of 250 objects pushed 250 times each, all with RGB-D video. First probabilistic meta-learning benchmark.

Graph Element Networks: adaptive, structured computation and memory
Ferran Alet , Adarsh K. Jeewajee, Maria Bauza, Alberto Rodriguez, Tomás Lozano-Pérez, Leslie Pack Kaelbling
ICML, 2019   (Long talk)
talk/ code

We learn to map functions to functions by combining graph networks and attention to build computational meshes and show this new framework can solve very diverse problems.

Modular meta-learning
Ferran Alet , Tomás Lozano-Pérez, Leslie Pack Kaelbling
CoRL, 2018  
video/ code

We propose to do meta-learning by training a set of neural networks to be composable, adapting to new tasks by composing modules in novel ways, similar to how we compose known words to express novel ideas.

Finding Frequent Entities in Continuous Data
Ferran Alet , Rohan Chitnis, Tomás Lozano-Pérez, Leslie Pack Kaelbling
IJCAI, 2018  
video/

People often find entities by clustering; we suggest that, instead, entities can be described as dense regions and propose a very simple algorithm for detecting them, with provable guarantees.

Robotic Pick-and-Place of Novel Objects in Clutter with Multi-Affordance Grasping and Cross-Domain Image Matching
Andy Zeng et al.
ICRA, 2018   (Best Systems Paper Award by Amazon Robotics)
talk/ project website

Description of the system for the Amazon Robotics Challenge 2017 competition, in which we won the stowing task.


He is a generous guy with a cool website.