Surface similarity parameter: A new machine learning loss metric for oscillatory spatio-temporal data.

Supervised machine learning approaches require the formulation of a loss functional to be minimized in the training phase. Sequential data are ubiquitous across many fields of research, and are often treated with Euclidean distance-based loss functions that were designed for tabular data. For smooth oscillatory data, those conventional approaches lack the ability to penalize amplitude, frequency and phase prediction errors at the same time, and tend to be biased towards amplitude errors. We introduce the surface similarity parameter (SSP) as a novel loss function that is especially useful for training machine learning models on smooth oscillatory sequences. Our extensive experiments on chaotic spatio-temporal dynamical systems indicate that the SSP is beneficial for shaping gradients, thereby accelerating the training process, reducing the final prediction error, increasing weight initialization robustness, and implementing a stronger regularization effect compared to using classical loss functions. The results indicate the potential of the novel loss metric particularly for highly complex and chaotic data, such as data stemming from the nonlinear two-dimensional Kuramoto-Sivashinsky equation and the linear propagation of dispersive surface gravity waves in fluids.

RIP2 regulates growth and differentiation of normal myoblasts and of rhabdomyosarcoma cells.

RIP2 is an important regulator of myoblast proliferation and differentiation. We have previously demonstrated that in the myoblast cell line C2C12 and in primary myoblasts, downregulation of rip2 gene expression is a prerequisite for differentiation. To further study the role of rip genes in myogenesis, we compared expression patterns of rip1-4 in two myoblast cell lines, C2C12 and C2F3, after the induction of differentiation. These two cell lines are derived from the same clonal origin, but differ with respect to their differentiation behaviour: specifically, the differentiation process is slower and more incomplete in C2F3 cells. When analyzing cells up to 4 days after the induction of differentiation, we found no downregulation of rip2 gene expression in C2F3 cells, which might be linked to the low differentiation potential of these cells. In addition, in contrast to C2C12 cells, the rip3 gene was not expressed in C2F3 cells. To further study the role of rip genes in the regulation of myoblast growth and differentiation, we analyzed expression patterns of rip1-4 in rhabdomyosarcoma cell lines. We found that in these cells, rip2 expression was not downregulated after the induction of differentiation. Furthermore, in contrast to normal myoblasts, they did not express the rip3 and rip4 genes. Thus, we focused on the functional role of RIP2 in rhabdomyosarcoma cells. Inhibition of rip2 gene expression in C2C12 and in rhabdomyosarcoma cells using specific siRNAs led to decreased proliferation and promoted the differentiation process of these cells. These data indicate that differential expression of rip genes can be associated with abnormal growth and differentiation behaviour of skeletal myoblasts.