Visual questions (VQs) can lead multiple people to respond with different answers rather than a single, agreed upon response. Moreover, the answers from a crowd can include different numbers of unique answers that arise with different relative frequencies. Such answer diversity arises for a variety of reasons including that VQs are subjective, difficult, or ambiguous. We propose a new problem of predicting the answer distribution that would be observed from a crowd for any given VQ; i.e., the number of unique answers and their relative frequencies. Our experiments confirm that the answer distribution can be predicted accurately for VQs asked by both blind and sighted people. We then propose a novel crowd-powered VQA system that uses the answer distribution predictions to reason about how many answers are needed to capture the diversity of possible human responses. Experiments demonstrate this proposed system accelerates capturing the diversity of answers with considerably less human effort than is required with a state-of-art system.

Constrained random simulation is becoming the mainstream methodology to verify system-wide properties in functional verification. It is a must to develop a high-throughput constrained random pattern generator, which is able to support arbitrary distribution. In this paper, we propose a novel range-splitting heuristic and a solution-density estimation technique to conquer the challenges of random pattern generators proposed in the recent literature. The solution densities can significantly increase by pruning infeasible subspaces. On the other hand, the estimated solution densities stored on a range-splitting tree statistically predict the distribution of solutions. Therefore, the generated patterns are ensured to meet the desired distribution with high throughput. Experimental results show that our framework achieves more than 10X speedup on average when compared to a commercial generator.

Constrained random simulation is becoming the mainstream methodology in functional verification. In order to achieve the verification closure, a high-throughput and evenly-distributed constrained random pattern generator has become a must. In this paper, we propose a novel Range-Splitting heuristic and a Solution-Density Estimation technique (RSSDE) to partition sample space. The chosen cutting planes target to prune more infeasible subspaces so that the solution densities in other subspaces increase correspondingly. In addition, with statistics-based analyses, the estimated solution densities precisely predict the distribution of solutions. The intermediate statistical information is recorded in a range-splitting tree (RS-tree). By top-down random walking on the RS-tree, random pattern generation produces evenly-distributed patterns with high throughput. Experimental results show that our framework guarantees evenly-distributed stimuli and achieves more than 10× speedup in average when compared to a state-of-the-art commercial generator.

Functional rectification in late design stages has been a crucial process in modern complex system design. This paper proposes a robust functional ECO engine, which applies SAT proof minimization and interpolation techniques to automate patch construction to make old implementation and golden specification functionally equivalent. The SAT proof minimization technique provides a sound and efficient way of fixing easy errors, and the interpolation technique provides a complete and robust way of fixing remaining errors. Experimental results show that our engine performs robustly to generate small patches in fixing various design rectification instances.

To approximate the exact peaks of continuous-time orthogonal frequency division multiplexing (OFDM) signals, four times oversampling is usually employed for discrete-time OFDM signals in various peak-to-average power ratio (PAPR) reduction schemes. Such an oversampling process would significantly increase the computational complexity, especially for a selected mapping scheme where a number of IFFT blocks may be used. In this paper, we present a low-complexity PAPR estimation scheme with two times oversampling based on "peak-search-and-partial-interpolation" (PSPI). As compared to the PAPR estimation method with four times oversampling, the proposed one achieves similar performance with only about half computational complexity. It is also shown that, with no significant increase in the computational complexity, the proposed PSPI scheme with two times oversampling has better PAPR estimation performance than the previous PSPI work without oversampling.

The high peak-to-average power ratio (PAPR) of the output signal for the transmitter is one of the main drawbacks of orthogonal frequency division multiplexing (OFDM) systems. To properly approximate the exact peaks of the continuous-time OFDM signal, L (usually ges 4) times oversampling is generally used for the discrete-time OFDM signal in many PAPR reduction schemes. However, the oversampling process increases the computational complexity. In this paper, we propose a low-complexity PAPR estimation method-peak-search-and- partial-interpolation" (PSPI) scheme-to efficiently estimate the PAPR value for the discrete-time OFDM signal. Instead of doing L times oversampling, the proposed PSPI method interpolates the 2(L-1) neighboring samples of L times oversampling for each searched peak of the original discrete-time OFDM signal. The simulation results illustrate that the proposed PSPI method can achieve well-approximated results as good as those of the scheme with four times oversampling, but with much lower computational complexity.