1: \begin{abstract}

2: 

3: This paper discusses a deterministic clustering approach to capacitated resource allocation problems. In particular, the Deterministic Annealing (DA) algorithm from the data-compression literature, which bears a distinct analogy to the phase transformation under annealing process in statistical physics, is adapted to address problems pertaining to clustering with several forms of size constraints. These constraints are addressed through appropriate modifications of the basic DA formulation by judiciously adjusting the free-energy function in the DA algorithm. At a given value of the annealing parameter, the iterations of the DA algorithm are of the form of a Descent Method, which motivate scaling principles for faster convergence.

4: 

5: \end{abstract}

6: