Our team VALENCIA IA4COVID has progressed to the 2nd phase of the Pandemic Response Challenge, organized by XPRIZE Foundation and supported by Cognizant. This is a $500K, a four-month challenge that focuses on the development of data-driven AI systems to predict COVID-19 infection rates and prescribe Intervention Plans (IPs) that regional governments, communities, and organizations can implement to minimize harm when reopening their economies. Our group is made up of more than twenty experts from the Universities and research centers of the Valencian Community and led by Dr. Nuria Oliver. We have all been working intensively since the beginning of the pandemic, altruistically and using the resources available to us in our respective institutions and with the occasional philanthropic collaboration of some companies.

Our model is among the three best models in the competition in MAE Mean Rank, leading in ASIA and in the top 5 of EUROPE in MAE per 100k habitants.

You can see our predictions here. The model has not been updated since its release on December 22nd.

Our paper Reversible Self-Replication of Spatio-Temporal Kerr Cavity Patterns has been recently accepted for publication in Physical Review Letters (PRL), with IF = 8.385, jointly with Salim B. Ivars, Yaroslav V. Kartashov, Lluis Torner, and Carles Milián. In this work, we uncover a novel and robust phenomenon that causes the gradual self-replication of spatiotemporal Kerr cavity patterns in cylindrical microresonators. These patterns are inherently synchronized multi-frequency combs. Under proper conditions, the axially-localized nature of the patterns leads to a fundamental drift instability that induces transitions amongst patterns with a different number of rows. Self-replications, thus, result in the stepwise addition or removal of individual combs along the cylinder’s axis. Transitions occur in a fully reversible and, consequently, deterministic way. The phenomenon puts forward a novel paradigm for Kerr frequency comb formation and reveals important insights into the physics of multi-dimensional nonlinear patterns.

Our recent paper Potential limitations in COVID-19 machine learning due to data source variability: A case study in the nCov2019 dataset has been accepted for publication in J Am Med Inform Assoc. (JAMIA, IF 4.112). We study whether the lack of representative coronavirus disease 2019 (COVID-19) data is a bottleneck for reliable and generalizable machine learning. Data sharing is insufficient without data quality, in which source variability plays an important role. We showcase and discuss potential biases from data source variability for COVID-19 machine learning. Our results are based in the publicly available nCov2019 dataset, including patient-level data from several countries. We aimed to the discovery and classification of severity subgroups using symptoms and comorbidities. We show that cases from the 2 countries with the highest prevalence were divided into separate subgroups with distinct severity manifestations. This variability can reduce the representativeness of training data with respect the model target populations and increase model complexity at risk of overfitting.

During the last months, I have been participating in the ANDI Challenge, together with my Ph.D. student Óscar Garibo. Since Albert Einstein provided a theoretical foundation for Robert Brown’s observation of the movement of particles within pollen grains suspended in water, significant deviations from the laws of Brownian motion have been uncovered in a variety of animate and inanimate systems, from biology to the stock market. Anomalous diffusion, as it has come to be called, is connected to non-equilibrium phenomena, flows of energy and information, and transport in living systems.

The challenge consists of three main tasks, each of them on 3 Dimensions:

• Task 1 – Inference of the anomalous diffusion exponent α.
• Task 2 – Classification of the diffusion model.
• Task 3 – Segmentation of trajectories.

We got the first position in Task 1 (1D) and the second position in Task 2 (1D). We also get the 3rd position in Task 2 (3d) and the 4th position in Task 2 (2D).