Some Recent Results on characterizations of Measures of Information Related to Coding
János Aczél · IEEE Transactions on Information Theory · 1978
Essential properties of entropies and other measures of information are arrived at on the basis of the noiseless coding theorems, source entropies, and Huffman codes (and also on the basis of forecasting and experiments). Conversely, characterization theorems are given based on these properties. Some of the most important are Shannon's inequality, boundedness on an interval, subadditivity, additivity, branching, and expansibility. Also entropies of mixed probabilistic and nonprobabilistic character and convexf-divergences are mentioned, among others. Some unsolved problems are stated.