The legacy of Jordan's canonical form on Poincaré's algebraic practices. This paper proposes a transversal overview on Henri Poincaré's early works (1878-1885). Our investigations start with a case study of a short note published by Poincaré on 1884 : "Sur les nombres complexes". In the perspective of today's mathematical disciplines - especially linear algebra -, this note seems completely isolated in Poincaré's works. This short paper actually exemplifies that the categories used today for describing some collective organizations of knowledge fail to grasp both the collective dimensions and individual specificity of Poincarés work. It also highlights the crucial and transversal role played in Poincaré's works by a specific algebraic practice of classification of linear groups by reducing the analytical representation of linear substitution to their Jordan's canonical forms. We then analyze in detail this algebraic practice as well as the roles it plays in Poincaré's works. We first provide a micro-historical analysis of Poincaré's appropriation of Jordan's approach to linear groups through the prism of the legacy of Hermite's works on algebraic forms between 1879 and 1881. This mixed legacy illuminates the interrelations between all the papers published by Poincaré between 1878 and 1885 ; especially between some researches on algebraic forms and the development of the theory of Fuchsian functions. Moreover, our investigation sheds new light on how the notion of group came to play a key role in Poincaré's approach. The present paper also offers a historical account of the statement by Jordan of his canonical form theorem. Further, we analyze how Poincaré transformed this theorem by appealing to Hermite's
