Value-Difference Based Exploration: Adaptive Control between Epsilon-Greedy and Softmax

240 citations recorded by the source corpus.

How researchers see this paper

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

Taming the Noise in Reinforcement Learning via Soft Updates (2015, methodology): “Soft-greedy policies have been used before for exploration [2, 36], but to our knowledge G-learning is the first TD-learning algorithm to explicitly use soft-greedy policies in its updates.”
DORA The Explorer: Directed Outreaching Reinforcement Action-Selection (2018, background): “Other options for evaluating exploration include recency and value difference (or error ) measures (Thrun, 1992; Tokic & Palm, 2011).”
G-Learning: Taming the Noise in Reinforcement Learning via Soft Updates (2015, methodology): “Soft-optimal policies have been used before for exploration [2, 37], but to our knowledge G-learning is the first TD-learning algorithm to explicitly use softoptimal policies in its updates.”
A Multi-Agent Reinforcement Learning-Based Data-Driven Method for Home Energy Management (2020, methodology): “To realize the exploration and exploitation, the ε-greedy policy (ε [0, 1]) [34] is adopted so the agent can either execute a random action form the set of available actions with probability ε or select an action whose current Q-value is maximum, with probability 1−ε.”