``What we need is more people who specialize in the impossible.''

-Theodore Roethke



    The laboratory specializes in ``applied applied math'', the art of doing important things faster than everyone else. We develope algorithms for processing big data from computational biology and mass spectrometry.

    Key contributions and research interests:

      Fast manipulations of tensors (``template-recursive iteration over tensors'' (TRIOT), a design pattern in C++11 for vectorizing operations over tensors as fast as Fortran, but with greater brevity and without knowing the dimensionality at compile time, ref: Serang, in review 2016).

      Fast algorithms for operating on discrete distributions (the probabilistic convolution tree, ref: Serang 2014, PLOS ONE and fast numeric max-convolution, refs: Serang 2015, Journal of Computational Biology; Pfeuffer and Serang 2016, Journal of Machine Learning Research).

      Dynamic programming for de novo analysis of glycoconjugates and small molecules (SweetSEQer, ref: Serang*, Froehlich* et al. Molecular & Cellular Proteomics 2013).

      Combinatorics of polyploid populations (SuperMASSA, ref: Serang, Molinari, & Garcia 2012, PLOS ONE).

      Nonparametric machine learning for evaluating discoveries using only empirical null samples (npCI, refs: Serang et al. 2012, Molecular & Cellular Proteomics; Serang et al. 2013, Journal of Proteome Research).



    SCICOMP is a graduate course in scientific computing created and taught at Universität Bremen in 2014 and continued at Freie Universität Berlin in 2015 and 2016, which covers topics in computer science and statistics with applications from biology. The course is designed top-down, starting with a problem and then deriving a variety of solutions from scratch.

    Topics include memoization, recurrence closed forms, string matching (sorting, hash tables, radix tries, and suffix tries), dynamic programming (e.g. Smith-Waterman and Needleman-Wunsch), Bayesian statistics (e.g. the envelope paradox), graphical models (HMMs, Viterbi, junction tree, belief propagation), FFT, and the probabilistic convolution tree.



    Stay Close, Little Ghost

    The ecstatic novel Stay Close, Little Ghost is a creative work of literature and mathematics that connects the quest for the holy grail of mathematical optimization, P=NP, to sexual and social dynamics, and then thrusts these forces against the universal desire to be loved as we are. What emerges is an intimate and defiant treatise on the nature of love and human imperfection. Critics have praised it as ``reminiscent of Murakami's finest moments'' (-The Next Best Book Blog), ``reminiscent of Camus' The Stranger or even Dostoyevsky's Crime and Punishment'' (-The Alternating Current Press), ``a love story of this generation'' (-The Lit Pub), and have written that ``the book pulses with the vibrant imagination of this unique outlook without ever becoming precious or pretentious. It's like a literary form of synaesthesia'' (-Amazon review).


    AT OCEAN is an ambitious novel, which blends cyberpunk and scientific themes (dangerous criminal and corporate organizations, cellular automata, chemistry and the search for stable, super-heavy isotopes, religiosity, ancient weapons that slice through steel, the distribution of prime numbers and their relationship to patterns hidden in classical art) into a meditation on our existential search for love and meaning in a universe with no apparent permanence. It has been praised as ``a terrific little surreal Sci-Fi conspiracy thriller'' (-The Digital Fix) and as ``like a dream'' and ``reminiscent of Borges'' (-Math Fiction).