Transfer of samples in batch reinforcement learning

By Alessandro Lazaric, Marcello Restelli, Andrea Bonarini. 2008

200 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Molecular Biology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Molecular Biology

Citations in the corpus

Learning by Playing - Solving Sparse Reward Tasks from Scratch (2018, neutral): “Our work is also connected to the broader literature on multi-task (reinforcement) learning (see e.g. Caruana (1997) for a general overview and Lazaric et al. (2008); Mehta et al. (2008) for RL applications) and work on reinforcement learning via options (Dietterich, 1998; Bacon et al., 2017;…”
Automatically Mapped Transfer between Reinforcement Learning Tasks via Three-Way Restricted Boltzmann Machines (2013, methodology): “[9] transfers samples in batch reinforcement learning using a compliance measure.”
Measuring the Distance Between Finite Markov Decision Processes (2016, background): “Some work just measures the distance between dynamics of two MDPs, such as reward functions or transition models, and requires that the state-action spaces are the same [2, 6, 19].”
Transferring Instances for Model-Based Reinforcement Learning (2008, background): “Recent empirical successes in a variety of RL domains [5–7] have shown that transfer can significantly increase an agent’s ability to learn quickly, even if agents in the two tasks have different available sensors or actions.”
Sharing Knowledge in Multi-Task Deep Reinforcement Learning (2020, unknown): “Similar considerations, which however does not use a Bayesian approach, are implicitly made in Taylor et al. (2007), Lazaric et al. (2008), and also in this work.”
Curriculum Learning for Reinforcement Learning Domains: A Framework and Survey (2020, background): “This knowledge can take the form of samples (Lazaric et al., 2008; Lazaric and Restelli, 2011), options (Soni and Singh, 2006), policies (Fernández et al.”
Autonomous Agents and Multi-agent Systems Manuscript No. Learning Potential Functions and Their Representations for Multi-task Reinforcement Learning (unknown): “Under a broader interpretation, initialization can also be performed by transferring source task experience samples to a batch-learning method in the target task [42].”
MT-Opt: Continuous Multi-Task Robotic Reinforcement Learning at Scale (2021, background): “Learning complex and composite skills has been addressed through hierarchical reinforcement learning with options [75, 7, 14], combining multiple sub-tasks [6, 16, 49, 26, 69, 89], reusing samples between tasks [39], relabeling experience in hindsight [3], introducing demonstrations [58, 81, 29, 41,”
Policy Transfer using Reward Shaping (2015, background): “transferring knowledge learned in a previous (source) task, has received a lot of attention in recent years, thanks to the publication of a few seminal papers that propose simple and intuitive techniques, achieving impressive improvements in learning [7, 11, 19].”
From N to N+1: Multiclass Transfer Incremental Learning (2013, methodology): “The paper is organized as follows: after a review of previous work (Section 2), we describe our setting (Section 3) and our algorithm (Section 4).”
Bayesian Reinforcement Learning (2010, unknown): “While some methods explicitly assume that the shared components are drawn from a common generative model (Wilson et al, 2007; Mehta et al, 2008; Lazaric and Ghavamzadeh, 2010), this assumption is more implicit in others (Taylor et al, 2007; Lazaric et al, 2008).”
Transfer Learning in Deep Reinforcement Learning: A Survey (2020, background): “A similar idea of mapping transition dynamics can be found in [121], which, however, requires a stronger assumption on the similarity of the transition probability and the state representations between the source and the target domains.”
Transfer in Reinforcement Learning: A Framework and a Survey (2012, methodology): “Let L be the number of source tasks, Lazaric et al (2008) propose an algorithm proposed which first collects N s samples for each source task K s = (S × A× S × R) N s and N t samples (with N t ≪ N s ) from the target task K t = (S × A × S × R) N t , and the hal-00772626, version 1 -10 Jan 2013…”
Online Transfer Learning in Reinforcement Learning Domains (2015, methodology): “While training, an agent pauses every 1. few episodes to perform at least 30 evaluation episodes and record its average performance — graphs show the performance of students when they are 1) not learning and 2) not receiving advice.”
Transfer Learning for User Adaptation in Spoken Dialogue Systems (2016, background): “This article does not deal with representation transfer, and has chosen instead to focus on transfer selection: source selection [14, 30] and instance selection [26].”