Paving the way for generalised systems with more effective and efficient AI
Are you ready for the thirty-ninth International Conference on Machine Learning (ICML 2022) happening from 17-23 July, 2022 at the Baltimore Convention Center in Maryland, USA? This hybrid event will bring together researchers working across artificial intelligence, data science, machine vision, computational biology, speech recognition, and more to showcase cutting-edge work in machine learning.
Excitingly, our research teams are presenting 30 papers, including 17 external collaborations. We are also proud sponsors of workshops and socials organized by our long-term partners LatinX, Black in AI, Queer in AI, and Women in Machine Learning. Let’s dive into a preview of our upcoming oral and spotlight presentations:
Effective reinforcement learning
Building generalised AI systems requires making reinforcement learning (RL) algorithms more effective. Our selected oral presentations showcase innovative approaches, such as applying generalised policy improvement (GPI) and exploring efficient exploration methods without the need for bonuses. We also introduce a method to augment RL agents with a memory-based retrieval process to enhance learning efficiency.
Progress in language models
Language is a powerful tool that shapes human communication and understanding. Our work focuses on building more efficient and effective language models through exploring unified scaling laws, retrieval methods, and adapting to new knowledge over time. We also address challenges in narrative generation to improve coherence and consistency in text generation.
Algorithmic reasoning
Neural algorithmic reasoning opens up possibilities for applying known algorithms to real-world problems. Our presentation includes benchmarks for evaluating neural networks on classical algorithms, an incremental learning algorithm for automated theorem proving, and a framework for constraint-based learned simulation, revolutionizing problem-solving in science and engineering.
Explore the full range of our work at ICML 2022 here.