### lable model? To th ture by means of neural networks. A neural network extracts a model from presented data by means of a learning algorithm. By lack of an explicit model, learning will automatically establish the dimensions (or hidden features) that are needed to fit the process at hand and augment the mathematical model for detection purposes. This allows us to achieve simultaneously the desired robustness and sensitivity, as discussed in the next section.

### Table 2. Overcoming knowledge

1998

"... In PAGE 8: ...integrates a diagnostic module for the prerequisite units in the presentation of the current unit. In Table2 some examples for the rules implemented in the system and the resulting presentation can be seen. Table 2: Example modifications of the default teaching strategies depending on learner characteristics Learner characteristic Teaching sequence, Presentation The learner did not work on the current unit and has specified no media preferences.... In PAGE 8: ... In Table 2 some examples for the rules implemented in the system and the resulting presentation can be seen. Table2 : Example modifications of the default teaching strategies depending on learner characteristics Learner characteristic Teaching sequence, Presentation The learner did not work on the current unit and has specified no media preferences. Default material sequences, no modifications.... In PAGE 19: ... Through a combination of different categories of sequencing algorithms specified by Frank [2] adaptive sequencing can be used in all three models of Salomon. Table2 gives an overview of adaptive sequencing algorithms used in ACE for different teaching goals. Adaptive sequencing can help to overcome knowledge deficits by adapting the complexity of learning material, compensate more general deficits through alternative learning sequences and adapt to individual preferences by selecting units or media.... ..."

Cited by 2

### Table 4. Technology-enhanced learning in online environments.

2005

"... In PAGE 19: ... 9) Use technology for online learning: Much has been written on the use of information technology in engineering education (see review [99]). Using a variety of these technologies, ranging from high-speed connectivity to course management systems, can assist that will facilitate learning engineering that cannot be done without technology? Table4 outlines examples that illustrate how technology permits the implementation of online paradigms that would be difficult without technology and how each of these examples affects quality, scale, and breadth. Activity made possible or improved through use of technology Technology implementation method Potential Effects: Quality, Scale, Breadth Student teams collaborate across multiple institutions Internet and multiple TCP/IP enabled technologies, both synchronous and asynchronous Quality, scale and breadth Robust game-playing simulations across institutions [100] Simulation software Quality Remote laboratories and instruments Remote control via the Web ... ..."

Cited by 4

### Table 6: Robustness of tournament sheets.

2005

"... In PAGE 16: ... That is, we will use one probability model to create a contrarian sheet, and see how it fares in ROI simulations performed under another model. Table6 gives the results, where rows correspond to true probability models and columns to the model used to derive the \optimal sheet. quot; This table can be read two difierent ways.... In PAGE 20: ... Table 10 contains the average ROI for this approach in its \predicted champion-only contrarian quot; columns. Comparing these results to Table6 gives an indication of how our predicted champion-only contrarian al- gorithm performs relative to the case where we know exactly how our opponents bet. In 58% of the cases, our algorithm obtained the same sheet as when we assume this knowledge, and therefore delivered an equal simulated ROI.... ..."

### Table 1: Enablers and Barriers to Effective Knowledge Capture for Project-Based Environments Enablers Barriers

"... In PAGE 7: ... A prerequisite for the development of a propositional framework of project-based learning dimensions was the development of a set of standardised descriptors of projects that could be usefully applied across different sectors. Illustrative examples from critical enablers and barriers in project probes are summarised in Table1 across seven key areas: project characteristics; project process characteristics; networking; project learning capture; organisational context; knowledge transfer and outcomes, which provide the basis for these standardised... ..."

### Table 1. Manual and machine representations Machine-learned

2001

"... In PAGE 6: ... The ontology learner [Maedche amp;Staab, 2000] applies this method straightforwardly for ontology learning from texts to support the knowledge engineer in the ontology acquisition environment. The main problem in applying ML algorithms for OL is that the knowledge bases constructed by the ML algorithms have a flat homogeneous structure, and very often have prepositional level representation (see Table1 ). Thus several efforts focus on improving ML algorithms in terms of ability to work with complicated structures.... ..."

Cited by 23

### Table 3 Performance with robust learning: (a) training set; (b) test set. Values in parentheses

in Robust Learning, Smoothing, and Parameter Tying on Syntactic Ambiguity Resolution Tung-Hui Chiang*

### Table 5. Robustness of algorithms

"... In PAGE 11: ....3. Sensitivity analysis We present a statistical analysis of many runs of our al- gorithms, each run with a different seed to the random num- ber generator. Table5 summarizes the results of this exper- iment. In row 1, we present the statistics for the Lazarus count of the total order found by Algorithm A1.... ..."

### Table 1: Conjunctive Rule Extraction Algorithm.

1994

"... In PAGE 4: ... Subset returns true if all of the in- stances that are covered by the rule are members of the given class, and false otherwise. Our algorithm for extracting conjunctive rules from trained neural networks is outlined in Table1 . It is an adaptation of the classical algorithm for PAC-learning monotone DNF expressions (Valiant, 1984).... In PAGE 5: ...i := randomly-select(vi1,...,vin) calculate the total input s to output unit if s then return e impose random order on all feature values /* consider the values in order */ for each value vij if changing feature ei apos;s value to vij increases s ei := vij if s then return e more directed Examples oracle for the general case. Note that the algorithm shown in Table1 employs a stopping criterion to determine when a set of extracted rules provides a su ciently-good model of a network. There are several reasonable criteria that could be used here.... In PAGE 7: ... Table 3 outlines the algorithm we use to extract M-of-N rules from trained networks. In the same man- ner as the algorithm presented in Table1 , the rst step is to learn a conjunctive rule using the instance sup- plied by the Examples oracle. The algorithm then makes this conjunction into a trivial M-of-N rule for which M is set to N.... ..."

Cited by 67