### Table 2. Classification results for multiple kernels (AUC)

### Table 3: Prescriptive Aesthetic Framework Based on Data and Aesthetic Literature

### Table 1. A taxonomy of image classification methods.

"... In PAGE 7: ... In general, image classification approaches can be grouped as supervised and unsupervised, or parametric and non- parametric, or hard and soft (fuzzy) classification, or per-pixel, subpixel, and per- field. Table1 provides brief descriptions of these categories. For the sake of convenience, this paper groups classification approaches as per-pixel, subpixel, per- field, contextual-based, knowledge-based, and a combination of multiple classifiers.... ..."

### Table 3, SIMD matrix multiplication kernels.

"... In PAGE 14: ... The M ops3 are based on op counts for the standard method (3). We compare the three algorithms in Table3 , where all calculations are performed in 64 bit precision. We present M ops gures and the percent of the total time spent in `OverHead apos;.... In PAGE 18: ... Also, the scaling step in Winograd apos;s algorithm can be performed as part of the prepro- cessing. This reduces the `Overhead apos; in Table3 , signi cantly. The use of Winograd as a computational kernel in Strassen apos;s algorithm, also slightly changes the error bound (14) to kEk (( n n0 )log2 12(9 8n2 0 + 23n0) ? 5n)ukAk kBk + O(u2) ; (15) Strassen apos;s algorithm will have l = log(k) levels of recursion and require approxi- mately k2:8 (kernel) matrix multiplications each of size N N.... ..."

### Table 3. Results of the SVM semantic classification for the learning and testing datasets using the properties (1) activity and (4) average size.

"... In PAGE 5: ... The best classification performance for the activity-based mapping were obtained with the linear kernel (Table 2). The reason for the poor classification with the RBF kernel can be explained with an analysis of the data presented in the Table3 , which shows the classification results for the learning and testing datasets using the properties only two properties: activity and average size. The reasons that only these two properties of the space have been chosen for the analysis are that the classification results are very similar to the ones obtained with the four properties, and it allows us to visualize the classification results in a 2D graph (Figure 4).... In PAGE 5: ... The reasons that only these two properties of the space have been chosen for the analysis are that the classification results are very similar to the ones obtained with the four properties, and it allows us to visualize the classification results in a 2D graph (Figure 4). As it can be noticed in the Table3 , the classification results for the learning dataset using the RBF kernel are better than the ones obtained with the linear kernel. But the same performance is not obtained with the testing dataset.... ..."

### Table 1: Nodes set for GP based classification. Functions

2006

"... In PAGE 2: ...1.2 Solution Candidate Representation Using Hy- brid Tree Structures The selection of the library functions is an important part of any GP modeling process [5] because this library should be able to represent a wide range of systems; Table1 gives an overview of the function set as well as the terminal nodes used for the classification experiments documented in this paper. As the reader can see in Table 1, mathematical func- tions and terminal nodes are used as well as boolean op- erators for building complex arithmetic expressions.... In PAGE 2: ....1.2 Solution Candidate Representation Using Hy- brid Tree Structures The selection of the library functions is an important part of any GP modeling process [5] because this library should be able to represent a wide range of systems; Table 1 gives an overview of the function set as well as the terminal nodes used for the classification experiments documented in this paper. As the reader can see in Table1 , mathematical func- tions and terminal nodes are used as well as boolean op- erators for building complex arithmetic expressions. There are in fact no structural restrictions for the use of boolean blocks in formulae; of course, [Then/Else] and boolean ex- pressions have to be connected to [IF] nodes, but there are no other restrictions regarding the use of boolean blocks within mathematical expressions.... ..."

Cited by 1

### Table 1: Nodes set for GP based classification. Functions

"... In PAGE 2: ...1.2 Solution Candidate Representation Using Hy- brid Tree Structures The selection of the library functions is an important part of any GP modeling process [5] because this library should be able to represent a wide range of systems; Table1 gives an overview of the function set as well as the terminal nodes used for the classification experiments documented in this paper. As the reader can see in Table 1, mathematical func- tions and terminal nodes are used as well as boolean op- erators for building complex arithmetic expressions.... In PAGE 2: ....1.2 Solution Candidate Representation Using Hy- brid Tree Structures The selection of the library functions is an important part of any GP modeling process [5] because this library should be able to represent a wide range of systems; Table 1 gives an overview of the function set as well as the terminal nodes used for the classification experiments documented in this paper. As the reader can see in Table1 , mathematical func- tions and terminal nodes are used as well as boolean op- erators for building complex arithmetic expressions. There are in fact no structural restrictions for the use of boolean blocks in formulae; of course, [Then/Else] and boolean ex- pressions have to be connected to [IF] nodes, but there are no other restrictions regarding the use of boolean blocks within mathematical expressions.... ..."

### Table 5: Results of Aesthetic Values (avs) of Mandarin Learning Web Pages by Using SDA

2008

"... In PAGE 6: ... This information included number of objects, width, height, area, difference between objects and frames, centre point of objects while the aesthetics values showed were Balance, Equilibrium, Symmetry, Sequence, Rhythm, Order and Complexity as well as the overall Aesthetics Value (OM). Figure 2: Execution Dialogue: Drag Objects on Interface Figure 3: Execution Dialogue: Click Count Aesthetic Value Button Figure 4: Object Model of Main Page (Group 1) with Objects Numbered Figure 5: Aesthetic Value of Main Page Interface Figure 6: Information of Objects Dragged and Aesthetic Values of Main Page (Group 1) As a concrete example of how we used this application for our research, Table5 showed the results of all of the aesthetics values of twelve web pages used in our research. These values were between 0 (the worst) and 1 (the best).... ..."

### Table 1. Common kernels

2002

"... In PAGE 2: ... Optimizing the SVM hyper-parameters is a model selection problem that needs adapting multiple parameter values at the same time. The parameters to tune are those that embed any kernel function as the parameter in an RBF kernel or the couple ( ; ) in case of KMOD kernel (see Table1 ). In addition, another parameter the optimization may consider is the trade-off parameter C which may have a strong effect on the SVM behavior for hard classification tasks.... ..."

Cited by 4