Unverified60% confidenceSolutionExact time
The ε-greedy policy selects actions randomly with probability ε and selects the action with the highest current Q-value with probability 1-ε; ε is typically set to 1.0 at the beginning of training and then linearly decayed to 0.01 or 0.1 over millions of frames.
2
Sources
60%
Confidence
Long-term
Relevance
8/6/2026
First Seen
Sources
Related Claims
UnverifiedIn heterogeneous asynchronous MMAB, if players are processed in decreasing order of their activation probabilities, the approximation ratio of the greedy algorithm can be improved to 1/√e.72% similarUnverifiedPolicies lacking domain randomization typically perform well only in the single simulator used for training and quickly fail when confronted with the physical uncertainties of a real robot.72% similarUnverifiedThe main result of the paper is that Q-Learning with zero initialization and constant step size can, with high probability, reduce the sup-norm error to within ε using Õ(H^4/(μ_min·ε^2)+h_star) state transitions (for 0<ε≤1).70% similarUnverifiedWith a fixed learning_rate, when batch_size scales linearly with the number of environments and num_epoch remains unchanged, the total displacement pushing the policy toward a new distribution in a single iteration may increase, manifesting as violent oscillation or even collapse of rewards in the early stages of training.69% similarUnverified采用域随机化（在训练早期就让ping在整个范围内随机采样）而非单向退火，可以让每个版本模型从一开始就见过各种延迟条件，避免后期崩溃风险69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/693804API
curl https://kongchang.com/api/v1/knowledge/claims/693804MCP
get_claim(id=693804)