• Publications
  • Influence
Control under communication constraints
TLDR
This paper forms a control problem with a communication channel connecting the sensor to the controller, and provides upper and lower bounds on the channel rate required to achieve different control objectives.
Stochastic linear control over a communication channel
TLDR
It is shown that optimal quadratic cost decomposes into two terms: A full knowledge cost and a sequential rate distortion cost.
AdaBelief Optimizer: Adapting Stepsizes by the Belief in Observed Gradients
TLDR
AdaBelief is proposed to simultaneously achieve three goals: fast convergence as in adaptive methods, good generalization as in SGD, and training stability; it outperforms other methods with fast convergence and high accuracy on image classification and language modeling.
Control over noisy channels
TLDR
This work presents a general necessary condition for observability and stabilizability for a large class of communication channels, and studies sufficiency conditions for Internet-like channels that suffer erasures.
The Capacity of Channels With Feedback
TLDR
A general feedback channel coding theorem based on Massey's concept of directed information is proved and the average cost optimality equation (ACOE) can be viewed as an implicit single-letter characterization of the capacity.
Loopy Belief Propogation and Gibbs Measures
TLDR
This work relates convergence of LBP to the existence of a weak limit for a sequence of Gibbs measures defined on the LBP's associated computation tree, and develops easily testable sufficient conditions for convergence.
On the Feedback Capacity of Power-Constrained Gaussian Noise Channels With Memory
TLDR
A new method for optimizing the channel inputs for achieving the Cover-Pombra block-length- n feedback capacity is developed by using a dynamic programming approach that decomposes the computation into n sequentially identical optimization problems where each stage involves optimizing O(L 2) variables.
Efficient and Dynamic Routing Topology Inference From End-to-End Measurements
TLDR
This paper proposes a general framework for designing topology inference algorithms based on additive metrics that can flexibly fuse information from multiple measurements to achieve better estimation accuracy and develops computationally efficient (polynomial-time) topology inferred algorithms.
Feedback capacity of finite-state machine channels
TLDR
The feedback capacity computation may be formulated as an average-reward-per-stage stochastic control problem, which is solved by dynamic programming and the value of the capacity is easily estimated using Markov chain Monte Carlo methods.
Optimal sequential vector quantization of Markov sources
The problem of optimal sequential vector quantization of Markov sources is cast as a stochastic control problem with partial observations and constraints, leading to useful existence results for
...
...