We study the model-based undiscounted reinforcement learning for partially observable Markov decision processes (POMDPs). The oracle we consider is optimal policy of POMDP with a known environment in terms average reward over an infinite horizon. propose algorithm this problem, building on spectral method-of-moments estimations hidden models, belief error control POMDPs and upper confidence bou...

Error modelling has played a major role in generating post-corrections of analogue to digital converters (ADC). Benefits by using parametric models for post-correction are that they requires less memory and that they are easier to identify for arbitrary signals. However, the parameters are estimated in two steps; firstly, the integral nonlinearity (INL) is estimated and secondly, the model para...

This paper presents a new method for scalarization of nonlinear multi-objective optimization problems. We introduce a special class of monotonically increasing sublinear scalarizing functions and show that the scalar optimization problem constructed by using these functions, enables to compute complete set of weakly efficient, efficient, and properly efficient solutions of multi-objective optim...

We describe a new framework of a sublinear expectation space and the related notions and results of distributions, independence. A new notion of G-distributions is introduced which generalizes our G-normal-distribution in the sense that mean-uncertainty can be also described. W present our new result of central limit theorem under sublinear expectation. This theorem can be also regarded as a ge...

We present a probabilistic construction of $\mathbb{R}^d$ -valued non-linear affine processes with jumps. Given set $\Theta$ parameters, we define family sublinear expectations on the Skorokhod space under which canonical process X is (sublinear) Markov generator. This yields tractable model for Knightian uncertainty expectation Markovian functional can be calculated via partial integro-differe...

this paper studies the perturbed klein-gordon equation by the aid of several methods of integrability. there are six forms of nonlinearity that are considered in this paper. the parameter domains are thus identified.

We consider randomized encodings (RE) that enable encoding a Turing machine Π and input x into its “randomized encoding” Π̂(x) in sublinear, or even polylogarithmic, time in the running-time of Π(x), independent of its output length. We refer to the former as sublinear RE and the latter as compact RE. For such efficient RE, the standard simulation-based notion of security is impossible, and we t...

It has recently been observed that sparse and compressible signals can be sketched using very few nonadaptive linear measurements in comparison with the length of the signal. This sketch can be viewed as an embedding of an entire class of compressible signals into a low-dimensional space. In particular, d-dimensional signals with m nonzero entries (m-sparse signals) can be embedded in O(m log d...

The output scores of a neural network classifier are converted to probabilities via normalizing over the scores of all competing categories. Computing this partition function, Z, is then linear in the number of categories, which is problematic as real-world problem sets continue to grow in categorical types, such as in visual object recognition or discriminative language modeling. We propose th...