Convergence Analysis on TD(0) with i.i.d. samples
1Convergence RateBhandari et al. (2018)
Evaluation results
| Method | Links | |
|---|---|---|
| Bhandari et al. (2018)Type of error=E[|v(X, θk) − v(X, θ∗)|2], Step size=O(K−1/2), Averaging=yes2026.06 | 1 | |
| Bhandari et al. (2018)Type of error=E[|θk − θ∗|2], Step size=O(k−1ω−1), Averaging=no2026.06 | 1 | |
| Srikant and Ying (2019)Type of error=E[|θk − θ∗|2], Step size=O(k−1ω−1), Averaging=no2026.06 | 1 | |
| Liu and Olshevsky (2021)Type of error=E[|v(X, θK) − v(X, θ∗)|2], Step size=O(K−1/2), Averaging=yes2026.06 | 1 | |
| Lee and Orabona (2025)Type of error=E[|v(X, θK) − v(X, θ∗)|2], Step size=O(K−1/2), Averaging=yes2026.06 | 1 | |
| Dalal et al. (2018)Type of error=E[|θk − θ∗|2], Step size=O(k−β), Averaging=no2026.06 | 1 | |
| Lakshminarayanan and Szepesvari (2018)Type of error=E[|θk − θ∗|2], Step size=O(1), Averaging=yes2026.06 | 1 | |
| Patil et al. (2023)Type of error=E[|θk − θ∗|2], Step size=O(1), Averaging=yes2026.06 | 1 | |
| Mitra (2024)Type of error=E[|v(X, θk) − v(X, θ∗)|2], Step size=O(1), Averaging=yes2026.06 | 1 | |
| Samsonov et al. (2024)Type of error=E[|v(X, θk) − v(X, θ∗)|2], Step size=O(1), Averaging=yes2026.06 | 1 | |
| PCTD(0) (Theorem 3.1)Type of error=E[|v(X, θk) − v(X, θ∗)|2], Step size=O(1), Averaging=yes2026.06 | 1 | |
| PCTD(0) (Theorem 3.3)Type of error=E[|v(X, θk) − v(X, θ∗)|2], Step size=O(1), Averaging=yes2026.06 | 1 |