Abstract. We introduce a method for making approximate Bayesian inference based on quantizing the hypothesis space and repartitioning it as observations become available. The method relies on approximating an optimal inference by using a probability distribution for quantized intervals of the unknown quantity, and by adjusting the intervals so as to obtain higher resolution in regions of higher probability, and vice versa. We repartition the hypothesis space adaptively with the aim of maximizing the mutual information between the approximate distribution and the exact distribution. It is shown that this approach is equivalent to maximizing the entropy of the approximate distribution, and we provide low-complexity algorithms for approximating multi-dimensional posterior distributions with tunable complexity/performance. The resulting quantized distribution for a one-dimensional case can be visualized as a histogram where each bar has equal area, but in general unequal width. The method can be used to provide adaptive quantization of arbitrary data sequences, or to approximate the posterior expectation of for instance some loss function by summing over a pre-specified number of terms.

Abstract—It is possible to improve the spectral efficiency of cellular systems by dynamically repartitioning the available bandwidth between different interfering and noninterfering subareas within and among sectors. Here, we investigate the problem of maximizing the expected system throughput in a two-sector area by dynamic bandwidth partitioning between two transmission modes: 1) reusing bandwidth in the low-interference area near either of the base stations and 2) using bandwidth for macrodiversity or single-base station transmission to avoid interference. Mode 2 is typically useful for giving users near the cell border higher bit rates. The suggested solution adapts the bandwidth partitioning to reflect local transmission capacities and bandwidth demands. Thus, it automatically decides on whether both transmission modes should be used and how much bandwidth should be used in each mode. The solution requires only limited knowledge of the future arrival rates and channel qualities and uses probability theory to find a robust bandwidth partitioning. Finally, we discuss access control and when to switch a user between the transmission modes to achieve high spectral efficiency and some minimum average quality of service. Index Terms—Diversity methods, land mobile radio cellular systems, scheduling. I.

On behalf of:

Ericsson in Göteborg. Uppsala University acted as lead partner. The objectives were to develop and to study algorithms and system design for future wireless packet data transmission beyond 3G. In particular, the objective of the program has been to develop and explore a new type of radio system based on adaptive transmission. Wireless IP was initiated as a multi-university research project within the SSF PCC program in 2000. It became a SSF IT program funded by 10 MSEK from mid-2002 to mid-2005, after which the program received a 2.5-year extension funded by 5 MSEK from mid 2006 to the end of 2008. During 2002-2003, the WIP program performed an intense design effort, which resulted in the design and evaluation of a downlink for an adaptive OFDM system. The work discussed and offered answers to the following questions: Can a balanced adaptive OFDM-based downlink be designed without too much overhead due to guard-bands and control signaling? What multiuser scheduling gains may be attained in such systems? What spectral efficiency may be expected by using adaptive OFDM-based downlinks