Perplexity

By
[Topic Model] Perplexity[Topic Model] Perplexity is a standard performance measure used to evaluate models of text data. It measures a model’s ability to generalise and predict new documents: the perplexity is an indication of the number of equally likely words that can occur at an arbitrary position in a document. A lower perplexity therefore indicates better generalisation. We calculate perplexity on the test corpus... is a standard performance measure used to evaluate models of text data. It measures a model’s ability to generalise and predict new documents: the perplexity[Topic Model] Perplexity is a standard performance measure used to evaluate models of text data. It measures a model’s ability to generalise and predict new documents: the perplexity is an indication of the number of equally likely words that can occur at an arbitrary position in a document. A lower perplexity therefore indicates better generalisation. We calculate perplexity on the test corpus... is an indication of
the number of equally likely words that can occur at an arbitrary position in a document. A lower perplexity[Topic Model] Perplexity is a standard performance measure used to evaluate models of text data. It measures a model’s ability to generalise and predict new documents: the perplexity is an indication of the number of equally likely words that can occur at an arbitrary position in a document. A lower perplexity therefore indicates better generalisation. We calculate perplexity on the test corpus... therefore indicates better generalisation. We calculate perplexity[Topic Model] Perplexity is a standard performance measure used to evaluate models of text data. It measures a model’s ability to generalise and predict new documents: the perplexity is an indication of the number of equally likely words that can occur at an arbitrary position in a document. A lower perplexity therefore indicates better generalisation. We calculate perplexity on the test corpus... on the test corpusIn linguistics, a corpus (plural corpora) or text corpus is a large and structured set of texts (nowadays usually electronically stored and processed). They are used to do statistical analysis and hypothesis testing, checking occurrences or validating linguistic rules within a specific language territory....
C∗ containing M∗ documents as follows:

p(C*) = exp \{ - { { \sum^{M^*}_{d=1} \log p(w_d) } \over { \sum^{M^*}_{d=1} N_d } } \}

A. De Waal, E. Barnard, Evaluating topic models with stability, 19th Annu. Symp. Pattern Recognit. Assoc. South Africa. (2008) 79–84. http://researchspace.csir.co.za/dspace/handle/10204/3016.