The first twenty-eight odes of Anacreon : in Greek and in English and in both languages, in prose as well as in verse : with variorum notes, a grammatical analysis and a lexicon /

Mode of access: Internet.

English of the XIVth century: Chaucer's Prologue and Knightes tale (with grammatical and philological notes) designed to serve as an introduction to the study of English literature,

Reissue of first edition, 1872.

A commentary, critical and grammatical, on St. Paul's Epistle to the Galatians, with a revised translation.

Mode of access: Internet.

A critical and grammatical commentary on St. Paul's Epistle to the Ephesians /

Mode of access: Internet.

A commentary, grammatical and exegetical, on the book of Job : with a translation, vol. 1 /

Covers Job, chapters 1-14.

A Latin grammar for schools : based on the principles and requirements of the Grammatical Society.

Mode of access: Internet.

The twenty-first book of Livy. With explanatory and grammatical notes and a vocabulary of proper names.

Mode of access: Internet.

The Kitab al-Maksur wal-mamdud. Being a treatise lexicographical and grammatical from manuscripts in Berlin, London [and] Paris.

Text in Arabic.

Vocabulary of the Yoruba language, to which are prefixed the Grammatical elements of the language.

I. English and Yoruba.--II. Yoruba and English.

Essai sur la langue et la littérature chinoises, avec cinq planches, contenant des textes chinois, accompagnés de traductions, de remarques et d'un commentaire littéraire et grammatical. Suivi de notes et d'une table alphabétique des mots chinois.

Mode of access: Internet.

A critical pronouncing dictionary, and expositor of the English language ... To which are prefixed principles of English pronunciation ... Likewise, rules to be observed by the natives of Scotland, Ireland and London ... and directions to foreigners ... The whole interspersed with observations, etymological, critical, and grammatical ... To which is annexed A key to the classical pronunciation of Greek, Latin, and Scripture proper names, &c.

"A key to the classical pronunciation ... New-York [n.d.]": (103 p. at end)

A concise system of grammatical punctuation, selected from various authors, for the use of students,

Mode of access: Internet.

Vocabulary of the Haussa language. Part I.--English and Haussa. Part II.--Haussa and English. And phrases, and specimens of translations. To which are prefixed, the grammatical elements of the Haussa language.

Mode of access: Internet.

The Oceanic languages : their grammatical structure, vocabulary, and origin,

"The work is mainly a grammar and dictionary of only one Oceanic language--that of Efate in the New Hebrides. The comparative portions refer only to a very few other Oceanic languages, and are quoted only to illustrate certain features in the grammar of the Efate."--Review by S. H. Ray in Man, VIII (1908) no. 40.

Using the EM Algorithm to Train Neural Networks: Misconceptions and a New Algorithm for Multiclass Classification

The expectation-maximization (EM) algorithm has been of considerable interest in recent years as the basis for various algorithms in application areas of neural networks such as pattern recognition. However, there exists some misconceptions concerning its application to neural networks. In this paper, we clarify these misconceptions and consider how the EM algorithm can be adopted to train multilayer perceptron (MLP) and mixture of experts (ME) networks in applications to multiclass classification. We identify some situations where the application of the EM algorithm to train MLP networks may be of limited value and discuss some ways of handling the difficulties. For ME networks, it is reported in the literature that networks trained by the EM algorithm using iteratively reweighted least squares (IRLS) algorithm in the inner loop of the M-step, often performed poorly in multiclass classification. However, we found that the convergence of the IRLS algorithm is stable and that the log likelihood is monotonic increasing when a learning rate smaller than one is adopted. Also, we propose the use of an expectation-conditional maximization (ECM) algorithm to train ME networks. Its performance is demonstrated to be superior to the IRLS algorithm on some simulated and real data sets.