Nonlinear Time Series Analysis of Electroencephalogram Tracings of Children with Autism
Methods of nonlinear time series analysis were compiled for use in the analysis of Electroencephalogram (EEG) tracings of children aged three to seven with varying degrees of autism in order to provide a quantitative means of diagnosing autism and determining its severity in a child. After determining the EEG leads to be used for analysis, the identified methods were coded and saved as functions on Scilab. To test the compiled program, a minimum of five EEG readings per cluster of children diagnosed with mild, moderate, severe and no autism will be obtained. The project was able to identify the mean, standard deviation, skewness, kurtosis and other higher order moments, the autocorrelation function, and the Fourier Series as the time-resolved statistical methods to be used for time series analysis. The nonlinear analysis methods identified include the use of the correlation integral, time-delay embedding and the Lorenz equations. One-way ANOVA testing will then be used on the numerical data obtained from the analysis to determine if a significant numerical differentiation has been obtained between the different clusters of EEG. This will provide a definitive way to medically diagnose autism, pinpointing children afflicted with the disorder and giving them proper treatment.\r Two copies of the "Abstract of Exhibit" (in English) should be sent to the National Taiwan Science Education Center or email to fung@mail.ntsec.gov.tw or yuonne@mail.ntsec.gov.tw before December 31, 2009.
佛手瓜卷鬚之向觸性及其參與蛋白質之探討
本研究利用佛手瓜的卷鬚探討向觸性的原理。本研究大致分為兩部份，一方面我們在卷鬚中發現了含量極為豐富的構造，此一螺旋狀構造分布於維管束中，且用雙縮?詴劑檢測後發現其含有蛋白質，且不具有運輸水分的功能；並發現此一構造的分布疏密，會影響到螺旋內側外側以及切割後片段泡溫水的彎曲方向。此外，在進行卷鬚蛋白質電泳的過程中，我們發現使用含尿素的緩衝液萃取蛋白質的效果最佳，1克的卷鬚乾重約可萃取到5毫克的蛋白質，且蛋白質總量會隨著卷鬚的成熟而遞減。利用軟體比對及質譜分析八個蛋白質點，得知此八點的蛋白質為：malate dehydrogenase, oxygen-evolving enhancer protein 1, oxyen-evolving enhancer protein 2, calreticulin, peroxidase, stromal 70 kDa heat shock-related protein, and AP2/ERF and B3 domain-containing transcription repressor。由此可知，向觸性為植物經過一連串訊號傳遞後，對外界刺激的順應。
The Use of Simulated Annealing in Predicting Solar Flare Activity
A program based on the simulated annealing (SA) algorithm was developed to allow for a longer-term prediction of the occurrence of solar flares, thus reducing the risks they may pose to astronauts, satellites, and ultimately the planet Earth. Sunspot data such as the solar cycle, the appearance of sunspots, the sun's magnetic field, the occurrence of solar flares and other factors that concern the sun were considered in the development of the algorithm. The program, coded in C++, included providing for an initial random state that pertained to the sun's state at a specified time. A neighborhood function was designed based on how sunspots are formed and how they disappear, and the probability function was designed using previous solar cycles to show relationships between the number of sunspots and the time it takes for the sun to reach solar maxima. Finally, the cooling function was designed as a representation of the time. Recorded sunspot counts obtained from the US National Geophysical Data Center (NGDC) during the solar cycle 22 from 1986 to 1996 was used as test data. The output was compared to the actual recorded solar events or the period. The program was found to be 66.7% accurate in predicting solar flares when compared on a smoothed trendline, while it exhibited a 50.5% accuracy when comparing day-to-day data. Since solar flare acitivty is predicted in terms of the trend of activity and intensity, improving daily data accuracy is superfluous. Smoothed trendline prediction accuracy may however be improved by altering the importance of each factor in predicting solar flare activity and by devising a different sunspot classification scheme for intensity and activity.