Readings

Introduction and Motivation

On the Measure of Intelligence Francois Chollet 2019

On the Binding Problem in Artificial Neural Networks Schmidhuber et al. 2020

A Spline Theory of Deep Learning ICML2018

Computational Power of RNNs ACL2018

Thinking Like Transformers NeurIPS2021

Basics of Meta-learning: Metric/Model/Optimisation-based

Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks ICML2017

Conditional Neural Processes ICML2018

Siamese Neural Networks for One-shot Image Recognition ICML2015

Matching Networks for One Shot Learning NeurIPS2016

Fast Context Adaptation via Meta-Learning ICML2019

Representation learning and disentangled representations

Disentangling by Factorising ICML2018

Neural Discrete Representation Learning NeurIPS2017

Disentangled Feature Representation for Few-shot Image Classification AAAI2022

Meta-learning with Disentanglement for Medical Image Segmentation MICCAI2021

MONET Unsupervised Scene Decomposition DeepMind2019

Neuro-symbolic Learning and Reasoning

One-shot Neuro-Deductive Program Synthesis IJCAI2019

DreamCoder: Wake-sleep Bayesian program learning NeurIPS2020

Neural Algorithmic Reasoning DeepMind 2021

Solving Visual Analogies using Neural Algorithmic Reasoning AAAI2022

Neural Abstract Reasoner NeurIPS2020

Continual Learning and Meta Reinforcement Learning

RL2- Fast Reinforcement Learning via Slow Reinforcement Learning ICLR2017

Learning to Continually Learn ECAI2020

Supermasks in Superposition NeurIPS2020

A Combinatorial Perspective on Transfer Learning NeurIPS2020

Recent papers in Meta-learning

Neural complexity measures NeurIPS2020

Rethinking Few-Shot Image Classification NeurIPS2020

The Role of Global Labels in Few-Shot Classification NeurIPS2021

Adaptive Risk Minimization- Learning to Adapt to Domain Shift NeurIPS2021

Meta Learning Backpropagation And Improving It NeurIPS2021

Meta-learning Deep Energy-based Memory Models ICLR2020

Project Papers (from below topics)

Meta-learning: Learning Novel Concepts by Imitating Drawings ICML2021

Continual Learning: Continual Learning via Local Module Composition NeurIPS2021

Continual Learning: DualNet Continual Learning Fast and Slow NeurIPS2021

Continual Learning: Optimizing Reusable Knowledge for Continual Learning via Metalearning NeurIPS2021

Reinforcement-learning: Disentangled Predictive Representation for Meta-RL NeurIPS2021

Reinforcement-learning: Markov State Abstractions for Deep Reinforcement Learning NeurIPS2021

Recent papers in Neuro-symbolic Learning

Concept Learning with Energy-Based Models ICLR2018

Neural-Symbolic Integration: A Compositional Perspective AAAI2021

Dynamic Inference with Neural Interpreters NeurIPS2021

Meta-learning & Disentanglement Theory; Causal Disentanglement

Meta-Learning by Adjusting Priors Based on Extended PAC-Bayes Theory ICML2018

When is Unsupervised Disentanglement Possible? NeurIPS2021

MetaCI: Meta-Learning for Causal Inference in a Heterogeneous Population NeurIPS2019

Applications of Meta-learning and Continual Learning

Continual Learning for Multivariate Time Series Tasks with Variable Input Dimensions ICDM2021

Meta-Learning for Few-Shot Time Series Classification CODS-COMAD2020

Meta-Learning for Black-box Optimization ECML-PKDD2019