Destacados

Cada año, un comité de expertos debe acometer una ardua tarea: de entre todas las publicaciones de ICREA, debe escoger unas cuantas que destaquen del resto. Es todo un reto: a veces los debates se acaloran, y siempre son difíciles, pero acaba saliendo una lista con las mejors publicaciones del año. No se concede ningún premio, y el único reconocimiento adicional es el honor de ser resaltado en la web de ICREA. Cada publicación tiene algo especial, ya sea una solución especialmente elegante, un éxito espectacular en los medios de comunicación o la simple fascinación por una idea del todo nueva. Independientemente de la razón, se trata de los mejores de los mejores y, como tales, nos complace compartirlos aquí.

LIST OF SCIENTIFIC HIGHLIGHTS

Format: yyyy

  • EGG: A tool to simulate language emergence in deep networks. (2019)

    Baroni, Marco (UPF)

    view details
    CLOSE

    EGG: A tool to simulate language emergence in deep networks.

    Deep artificial neural networks are increasingly pervasive in our daily lives, helping us in everyday tasks such as tagging pictures or translating from one language to the other. However, these computational systems are not endowed with the ability to communicate with each other and with us, which makes them rather inflexible and opaque tools. Marco Baroni and his colleagues are interested in the question: what happens if we let a community of artificial deep network "invent" their own language in order to solve a task together. In order to encourage research in this interdisciplinary area, involving artificial intelligence, linguistics and cognitive science, in 2020 the team open-sourced the EGG toolkit. Using the toolkit (currently starred more than 100 times on GitHub, and presented at the prestigious EMNLP conference) Baroni and colleagues were able to highlight several interesting properties of the emergent system developed by deep networks to communicate. For example, the networks are not subject to the same energy saving constraints that shape the communication systems of humans and animals. Consequently, they might associate very long forms, such as "fjksjkgjrgjkrgksfkeeeeeeff", to very frequent words (e.g., those meaning "the" or "it"). Conversely, neural networks can invent very clever ways to communicate about the world (e.g., by referring to the relative intensity of different pixels) that lead to languages that are completely obscure for us, but allow extremely efficient information transmission. The research goal set for 2020 is to find a common ground between the language spoken by deep networks and the one spoken by people!

  • Reducing Child Deaths in Low-Income Countries. (2019)

    Bassat Orellana, Quique (ISGlobal)

    view details
    CLOSE

    Reducing Child Deaths in Low-Income Countries.

    A simple algorithm could contribute to reducing the high mortality among newborns and babies in the month following their hospital discharge.

    In the last 25 years, the reduction in mortality of children under five years of age has been remarkable but insufficient. In low-income countries, children are at increased risk of dying following hospitalization, regardless of their illness, with an estimated risk ranging between 3 and 13% in the month following discharge. The challenge, therefore, is to identify those children at higher risk in order to follow them up closely after discharge, and thereby avoid a considerable number of pediatric deaths.

    We conducted a retrospective study analyzing data from more than 20,000 pediatric hospital admissions over almost 20 years, in the district hospital of Manhiça, a semi-rural area in Southern Mozambique where almost half of the population is under 15 years of age. We determined mortality during the first, second and third month after hospital discharge, and looked for indicators that would allow to identify and eventually target children at higher risk of dying. 

    The results show that the average mortality after discharge is high (3.6%), a figure even higher than the documented intra-hospital mortality. Half of the post-discharge deaths occur within the first 30 days. The risk is highest in babies under 3 months of age and decreases progressively with age. The study also identifies a series of clinical parameters (malnutrition, diarrhea, clinical pneumonia, etc.) that allow to identify those children at highest mortality risk. Using all or some of these variables, the team used a series of predictive models capable of identifying up to 80% of children at risk of dying after discharge.

    The children thus identified could benefit from a close follow-up during the first 30 days by community health workers, or receive preventive antimicrobial therapies. If these simple models, based on easy-to-obtain parameters like those used in this study, are validated in other contexts, they could represent a valuable tool to save neonatal and infant lives in countries with a high burden of child mortality.

  • Freshwater biodiversity is declining at an alarming rate in Europe and Central Asia (2019)

    Brucet, Sandra (UVIC)

    view details
    CLOSE

    Freshwater biodiversity is declining at an alarming rate in Europe and Central Asia

    Freshwater systems are the most threatened ecosystem type in Europe and Central Asia region, with the quantity and quality of habitats and abundance of many species rapidly declining. This is the conclusion from our review paper which is part of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES) report on Europe and Central Asia.
    In the study we also show that only about half of the EU’s rivers and lakes achieved good ecological status in 2015 (as defined by the Water Framework Directive in terms of the quality of the biological community), and many lakes, ponds, and streams are disappearing as a consequence of agricultural intensification and ineffective irrigation and urbanisation, combined with climate change. The situation regarding freshwater biodiversity remains highly critical in Europe and Central Asia as many species remain threatened with extinction, including >50% of known species for some groups (e.g. molluscs, amphibians). 
    The reasons for the decline in freshwater biodiversity are the destruction or modification of their habitat, including water abstraction, which affects ∼89% of all amphibian threatened species and ∼26% of threatened freshwater invertebrate species. Of particular concern is the lack of data for freshwater invertebrates. Current status is available for only a minority of species, and the impact of alien invasive species is often unknown, especially in Central Asia. 
    Based on current freshwater biodiversity trends, it is highly unlikely that Europe and Central Asia will achieve either the respective Aichi biodiversity targets by 2020 (i.e., targets, 2–4,6–12,14) or Target 1 of the Biodiversity Strategy.

     

  • Strong coupling of artificial atoms mediated by waveguide photons (2019)

    Chang, Darrick (ICFO)

    view details
    CLOSE

    Strong coupling of artificial atoms mediated by waveguide photons

    Typically, excited atoms in open environments will exhibit dissipative and irreversible dynamics through the spontaneous emission of photons. However, by precisely positioning artificial atoms in the form of superconducting transmon qubits along a one-dimensional waveguide, we experimentally observed that the same photons that typically give rise to dissipation can also produce strong, coherent dynamics. This phenomenon is produced due to dynamical exchange between a single artificial atom and an emergent entangled state of the array. The ability to coax coherent dynamics from such a nominally dissipative system provides a route toward diverse applications like the generation of exotic quantum many-body states of light, quantum computing in decoherence-free subspaces, and multi-photon quantum state synthesis.

  • Malaria parasites have alternative pathways for sexual conversion (2019)

    Cortés Closas, Alfred (ISGlobal)

    view details
    CLOSE

    Malaria parasites have alternative pathways for sexual conversion

    The fundamental characteristics of the life cycle of malaria parasites were discovered over a century ago, but some specific steps are not well understood yet. One such step is the transition from asexual blood stages into sexual forms termed gametocytes, necessary for human to mosquito transmission. The results presented in this article modify the previous textbook view of this part of the malaria life cycle.

     

    All malaria clinical symptoms in humans are produced by repeated cycles of exponential asexual parasite growth in the blood. However, transmission of the disease from one human to another via a mosquito vector requires that some of the asexual parasites convert into non-replicating sexual forms termed gametocytes. Gametocytes are the only malaria parasite stages able to infect mosquitoes. For many years, the widely accepted model was that at each cycle of asexual growth a small subset of the parasites commits to sexual conversion, and then they must go through an additional round of multiplication as sexually-committed forms before they actually convert into gametocytes. However, a detailed characterization of the process was previously not possible because molecular markers for sexually-committed forms were not available.

     

    To address this gap of knowledge, we took advantage of our previous identification (in collaboration with other teams) of the PfAP2-G transcription factor as the master regulator of sexual conversion in malaria parasites. Using PfAP2-G as a marker for sexually committed parasites, we found that the additional round of multiplication after sexual commitment is not an obligate step. Thus, parasites can follow two alternative pathways after committing to sexual development: multiplying for one additional cycle before actual conversion into sexual forms (which we termed next cycle conversion route), or converting directly after commitment (same cycle conversion route). The availability of the two alternatives routes provides plasticity to the process, to either increase the output of sexual forms or ensure rapid conversion that may enable survival under harsh conditions.

  • Cells with double number of chromosomes can generate three cells during mouse embryo development. (2019)

    Cosma, Maria Pia (CRG)

    view details
    CLOSE

    Cells with double number of chromosomes can generate three cells during mouse embryo development.

    Cells with high ploidy content, i.e with increased number of chromosomes, are common in mammalian adult tissues. Cell-to-cell fusion generates polyploid cells during mammalian development and tissue regeneration. However, whether increased ploidy can be occasionally tolerated during embryogenesis still remains largely unknown. We showed that tetraploid cells, when injected in a recipient mouse embryo, can generate diploid cells upon tripolar mitosis reduction. The generated diploid cells include the correct number of chromosomes and form part of the adult tissues in mouse. Overall, we discovered an unexpected process of controlled genome reduction generating diploid cells during mouse embryo development.

Pages