Papers by Maddalena Amendola
Latent Assimilation: assimilating data in a latent space of a surrogate model
Formulation of a new methodology that combines machine learning and data assimilation techniques.... more Formulation of a new methodology that combines machine learning and data assimilation techniques. The methodology consists in using an Autoencoder to reduce the size of the input. In the latent space, a recurrent neural network (LSTM) is used as a surrogate for a dynamic system. The accuracy of the model is improved by using the Kalman Filter in the latent space which incorporates data (observation) collected by sensors, producing the updated state. The updated state is then reported in the original physical space by the decoder. The methodology was applied to a real test case
Social Search research deals with studying methodologies exploiting social information to better ... more Social Search research deals with studying methodologies exploiting social information to better satisfy user information needs in Online Social Media while simplifying the search effort and consequently reducing the time spent and the computational resources utilized. Starting from previous studies, in this work, we analyze the current state of the art of the Social Search area, proposing a new taxonomy and highlighting current limitations and open research directions. We divide the Social Search area into three subcategories, where the social aspect plays a pivotal role: Social Question&Answering, Social Content Search, and Social Collaborative Search. For each subcategory, we present the key concepts and selected representative approaches in the literature in greater detail. We found that, up to now, a large body of studies model users' preferences and their relations by simply combining social features made available by social platforms. It paves the way for significant research to exploit more structured information about users' social profiles and behaviors (as they can be inferred from data available on social platforms) to optimize their information needs further.
ArXiv, 2020
There is an urgent need to build models to tackle Indoor Air Quality issue. Since the model shoul... more There is an urgent need to build models to tackle Indoor Air Quality issue. Since the model should be accurate and fast, Reduced Order Modelling technique is used to reduce the dimensionality of the problem. The accuracy of the model, that represent a dynamic system, is improved integrating real data coming from sensors using Data Assimilation techniques. In this paper, we formulate a new methodology called Latent Assimilation that combines Data Assimilation and Machine Learning. We use a Convolutional neural network to reduce the dimensionality of the problem, a Long-Short-TermMemory to build a surrogate model of the dynamic system and an Optimal Interpolated Kalman Filter to incorporate real data. Experimental results are provided for CO2 concentration within an indoor space. This methodology can be used for example to predict in real-time the load of virus, such as the SARS-COV-2, in the air by linking it to the concentration of CO2.
Data Assimilation in the Latent Space of a Convolutional Autoencoder
Uploads
Papers by Maddalena Amendola