Recency-weighted Markovian inference
arXiv:1711.03038v12 citations
Originality Synthesis-oriented
AI Analysis
This work addresses computational efficiency challenges in probabilistic modeling for researchers in machine learning and statistics, but appears incremental as it builds on existing Markovian frameworks.
The authors tackled the problem of approximating high-order Markov latent state space models with decaying mixture distributions, and developed a sampling algorithm that achieves fixed time and memory costs.
We describe a Markov latent state space (MLSS) model, where the latent state distribution is a decaying mixture over multiple past states. We present a simple sampling algorithm that allows to approximate such high-order MLSS with fixed time and memory costs.