Research map / Optimization / Optimizers
Infinity Learning: Learning Markov Chains from Aggregate Steady-State Observations
Jianfei Gao, Mohamed A. Zahran, Amit Sheoran, Sonia Fahmy, Bruno Ribeiro
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use ∞-SGD instead of DC-BPTT | parametric CTMC learning · synthetic queue simulations, heavy-load failure states | MAPE/100: DC-BPTT and ∞-SGD values are reported in Table [MAPE] | Improved |
| Use ∞-SGD instead of DC-BPTT | parametric CTMC learning · synthetic queue simulations, heavy-load failure states | MSE: DC-BPTT and ∞-SGD values are reported in Table [MSE] | Improved |
| Use the Upper Triangular model instead of the M/M/1/K model | ∞-SGD · VoLTE testbed, heavy-load call-drop probability | MAPE: about 80% to 100% vs 449% | Improved |
| Use DC-BPTT t*=16 instead of t*=128 (ablation) | DC-BPTT · M/M/1/K fast-mix | MAPE/100: 2.04 × 10^{-1} ± 2.86 × 10^{-4} vs 1.32 ± 4.94 × 10^{-3} | Improved |
| Use DC-BPTT t*=16 instead of t*=128 (ablation) | DC-BPTT · M/M/1/K fast-mix | MSE: 6.81 × 10^{-3} ± 1.99 × 10^{-5} vs 2.45 × 10^{-1} ± 1.84 × 10^{-3} | Improved |
| Use the strict M/M/1/K parameterization rather than the more flexible Upper Triangular parameterization (ablation) | ∞-SGD · VoLTE testbed, heavy-load call-drop probability | MAPE: about 80% to 100% vs 449% | Hurt |
| Allow additional non-parametric transition-rate parameters with α=0.1 (ablation) | parametric CTMC model · queue-loss extrapolation | extrapolation error | Hurt |
| Use some additional model flexibility with α≈1 (ablation) | parametric CTMC model · queue-loss extrapolation | extrapolation performance | No clear effect |
| Relax the parametric model substantially (ablation) | parametric CTMC model · queue-loss extrapolation | extrapolation performance | Hurt |
Closest papers by meaning
- Learning Multiple Markov Chains via Adaptive Allocation (NeurIPS 2019 · similarity 0.90)
- On Markov Chain Gradient Descent (NeurIPS 2018 · similarity 0.89)
- Stochastic Gradient Descent under Markovian Sampling Schemes (ICML 2023 · similarity 0.87)
- Deep Learning for Computing Convergence Rates of Markov Chains (NeurIPS 2024 · similarity 0.87)
- Laplace Transform Based Low-Complexity Learning of Continuous Markov Semigroups (ICML 2025 · similarity 0.86)
- Learning Extrapolative Sequence Transformations from Markov Chains (ICML 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
