### Table 3: A 15-bit error-correcting output code for a ten-class problem.

1995

"... In PAGE 4: ... Table3 shows a 15-bit error-correcting code for the digit-recognition task. Each class is represented byacodeword drawn from an error-correcting code.... In PAGE 4: ... If we make only b d,1 2 c errors, the nearest codeword will still be the correct codeword. #28The code of Table3 has minimum Hamming distance seven and hence it can correct errors in any three bit positions.#29 The Hamming distance between anytwo codewords in the one- per-class code is two, so the one-per-class encoding of the k output classes cannot correct any errors.... In PAGE 7: ... 2.3 Error-Correcting Code Design We de#0Cne an error-correcting code to be a matrix of binary values such as the matrix shown in Table3 . The length of a code is the number of columns in the code.... ..."

Cited by 353

### Table 1: 15-bit Error-correcting output code for a 10-class problem Code Word (si) Class i

1995

"... In PAGE 2: ...) We will call each string si the codeword for class ci. Table1 shows an example of a set of 10 such binary strings. The Hamming distance between each pair of these strings is always at least 7 bits.... ..."

Cited by 132

### Table 2. An Error Correcting Codes (ECC) Matrix

"... In PAGE 7: ... Dynamic Fusion Problem Viewed as a Factorial Hidden Markov Model (FHMM) We also define the ECC matrix ECC = [emn] as the diagnostic matrix (D-matrix), which represents the full-order dependency among failure sources and classifiers. Table2 shows an ECC matrix as an illustrative example (8 failure sources and 4 classifiers). Table 2.... ..."

### Table 1: Error correcting code for n = 8 classes

"... In PAGE 2: ... It is coded in a distributed representation of l outputs from all classi ers. Table1 shows an example of an error correcting code for n =8 classes with l =5codewords. The code words are the columns of the table.... ..."

### Table 3.3: Error-Correction Output Coding: (a) Hand crafted partial models, (b) Fully expanded bitvectors.

in OF

### Table 3: Error-Correction Interpretation: The standard table of a simple 1 error correcting code. Data Hiding Interpretation: Look-up table for decoding three hidden bits (syndrome).

2004

"... In PAGE 49: ... The codewords and the syndrome information is given in Table 3. The data hiding encoder works as follows: Given a hidden bit combination and an input word, the encoder searches the row of the Table3 whose coset label is the same as the given hidden bit combination. The word with the minimum distance to the given word is the output of the hidden data encoder.... ..."

### Table Page 1. Error-Correcting Codes ............................... 11

### Table 6: Comparison of output encoding methods for the Separate Phoneme/Stress approach. L is the length of the error-correcting code, and d is the minimum Hamming distance between each pair of codewords.

1999

"... In PAGE 7: ... We applied BCH code design methods (Bose amp; Ray-Chaudhuri, 1960; Hocquenghem, 1959; Lin amp; Costello, 1983) to design good error correcting codes of various lengths for both the separate and combined phoneme/stress con gurations. Table6 shows the results of the base con guration and the three alternative output encodings for the Separate Phoneme/Stress approach. Table 7 shows corresponding results for the Combined Phoneme/Stress approach.... ..."

Cited by 21