课程名称︰ 生物资讯学演算法 期末考 课程性质︰ 资讯系选修 课程教师︰ 欧阳彦正 开课学院： 电机资讯学院 开课系所︰ 资讯系 考试日期（年月日）︰ 2006/1/13 考试时限（分钟）： 是否需发放奖励金： yes （如未明确表示，则不予发放） 试题 : (20%)At a sub-root during the construction of a decision tree, the software needs to determine whether the acctivity of a gene could be exploited to predict the category that a person should belong to based on his/her weight. The following table gives the distribution of the samples at the sub-root. Assume that the criterion to prevent overfitting is that the statistical confidence of claiming the independence between the attribute and the decision is over 95%. Based on this criterion, can you tell whether overfitting could occur if the activity of the gene is applied to make the prediction? (20%)A biochemist wnats to test his hypothesis that the activities of gene1 and gene2 together determine whether a person suffers a disease. Following is the microarray data that the biochemist has obtained in as experiment. If the biochemist employs the univariate approach, can the biochemist figure out the influences of these two genes correctly? Assume that a confidence level of 95% is normally required ofr making a claim. Given that \sigma^2 for gene1 and gene2 are estimated to be 16.63 and 21.38, respectively. (20%)given the plot of the dataset in problem 2 as follos. Is this dataset linearly separable? if yes, can you figure out a linear decision function in th form of sgn(v) = sgn[ax+by+c], where v is the feature vector of a new sample and a, b, c are real numbers? (20%)Please describe the concept of PSI-BLAST by explaining how it incorporates sequence alignment(BLAST) with multiple sequence alignment (MSA) and position specific score matrix to fine remote homologues. (20%)Assume that you are requested to implement a software program that measures the structural similarity between a target protein and the reference protein. Furthermore, assume that you have found a software package that can provide you with the major principal components of a 3-dimensional object. How will you design the structural similarity analysis software with the Fast Fourier Ttansform algorithm?_ --

※ 发信站: 批踢踢实业坊(ptt.cc) ◆ From: 61.228.41.12