-
vEEGNet: learning latent representations to reconstruct EEG raw data via variational autoencoders
Authors:
Alberto Zancanaro,
Giulia Cisotto,
Italo Zoppis,
Sara Lucia Manzoni
Abstract:
Electroencephalografic (EEG) data are complex multi-dimensional time-series that are very useful in many applications, from diagnostics to driving brain-computer interface systems. Their classification is still a challenging task, due to the inherent within- and between-subject variability and their low signal-to-noise ratio. On the other hand, the reconstruction of raw EEG data is even more diffi…
▽ More
Electroencephalografic (EEG) data are complex multi-dimensional time-series that are very useful in many applications, from diagnostics to driving brain-computer interface systems. Their classification is still a challenging task, due to the inherent within- and between-subject variability and their low signal-to-noise ratio. On the other hand, the reconstruction of raw EEG data is even more difficult because of the high temporal resolution of these signals. Recent literature has proposed numerous machine and deep learning models that could classify, e.g., different types of movements, with an accuracy in the range 70% to 80% (with 4 classes). On the other hand, a limited number of works targeted the reconstruction problem, with very limited results. In this work, we propose vEEGNet, a DL architecture with two modules, i.e., an unsupervised module based on variational autoencoders to extract a latent representation of the data, and a supervised module based on a feed-forward neural network to classify different movements. To build the encoder and the decoder of VAE we exploited the well-known EEGNet network. We implemented two slightly different architectures of vEEGNet, thus showing state-of-the-art classification performance, and the ability to reconstruct both low-frequency and middle-range components of the raw EEG. Although preliminary, this work is promising as we found out that the low-frequency reconstructed signals are consistent with the so-called motor-related cortical potentials, well-known motor-related EEG patterns and we could improve over previous literature by reconstructing faster EEG components, too. Further investigations are needed to explore the potentialities of vEEGNet in reconstructing the full EEG data, generating new samples, and studying the relationship between classification and reconstruction performance.
△ Less
Submitted 16 November, 2023;
originally announced December 2023.
-
hvEEGNet: exploiting hierarchical VAEs on EEG data for neuroscience applications
Authors:
Giulia Cisotto,
Alberto Zancanaro,
Italo F. Zoppis,
Sara L. Manzoni
Abstract:
With the recent success of artificial intelligence in neuroscience, a number of deep learning (DL) models were proposed for classification, anomaly detection, and pattern recognition tasks in electroencephalography (EEG). EEG is a multi-channel time-series that provides information about the individual brain activity for diagnostics, neuro-rehabilitation, and other applications (including emotions…
▽ More
With the recent success of artificial intelligence in neuroscience, a number of deep learning (DL) models were proposed for classification, anomaly detection, and pattern recognition tasks in electroencephalography (EEG). EEG is a multi-channel time-series that provides information about the individual brain activity for diagnostics, neuro-rehabilitation, and other applications (including emotions recognition). Two main issues challenge the existing DL-based modeling methods for EEG: the high variability between subjects and the low signal-to-noise ratio making it difficult to ensure a good quality in the EEG data. In this paper, we propose two variational autoencoder models, namely vEEGNet-ver3 and hvEEGNet, to target the problem of high-fidelity EEG reconstruction. We properly designed their architectures using the blocks of the well-known EEGNet as the encoder, and proposed a loss function based on dynamic time warping. We tested the models on the public Dataset 2a - BCI Competition IV, where EEG was collected from 9 subjects and 22 channels. hvEEGNet was found to reconstruct the EEG data with very high-fidelity, outperforming most previous solutions (including our vEEGNet-ver3 ). Furthermore, this was consistent across all subjects. Interestingly, hvEEGNet made it possible to discover that this popular dataset includes a number of corrupted EEG recordings that might have influenced previous literature results. We also investigated the training behaviour of our models and related it with the quality and the size of the input EEG dataset, aiming at opening a new research debate on this relationship. In the future, hvEEGNet could be used as anomaly (e.g., artefact) detector in large EEG datasets to support the domain experts, but also the latent representations it provides could be used in other classification problems and EEG data generation.
△ Less
Submitted 20 November, 2023;
originally announced December 2023.
-
ACTA: A Mobile-Health Solution for Integrated Nudge-Neurofeedback Training for Senior Citizens
Authors:
Giulia Cisotto,
Andrea Trentini,
Italo Zoppis,
Alessio Zanga,
Sara Manzoni,
Giada Pietrabissa,
Anna Guerrini Usubini,
Gianluca Castelnuovo
Abstract:
As the worldwide population gets increasingly aged, in-home telemedicine and mobile-health solutions represent promising services to promote active and independent aging and to contribute to a paradigm shift towards patient-centric healthcare. In this work, we present ACTA (Advanced Cognitive Training for Aging), a prototype mobile-health solution to provide advanced cognitive training for senior…
▽ More
As the worldwide population gets increasingly aged, in-home telemedicine and mobile-health solutions represent promising services to promote active and independent aging and to contribute to a paradigm shift towards patient-centric healthcare. In this work, we present ACTA (Advanced Cognitive Training for Aging), a prototype mobile-health solution to provide advanced cognitive training for senior citizens with mild cognitive impairments. We disclose here the conceptualization of ACTA as the integration of two promising rehabilitation strategies: the "Nudge theory", from the cognitive domain, and the neurofeedback, from the neuroscience domain. Moreover, in ACTA we exploit the most advanced machine learning techniques to deliver customized and fully adaptive support to the elderly, while training in an ecological environment. ACTA represents the next-step beyond SENIOR, an earlier mobile-health project for cognitive training based on Nudge theory, currently ongoing in Lombardy Region. Beyond SENIOR, ACTA represents a highly-usable, accessible, low-cost, new-generation mobile-health solution to promote independent aging and effective motor-cognitive training support, while empowering the elderly in their own aging.
△ Less
Submitted 17 February, 2021;
originally announced February 2021.
-
Comparison of Attention-based Deep Learning Models for EEG Classification
Authors:
Giulia Cisotto,
Alessio Zanga,
Joanna Chlebus,
Italo Zoppis,
Sara Manzoni,
Urszula Markowska-Kaczmar
Abstract:
Objective: To evaluate the impact on Electroencephalography (EEG) classification of different kinds of attention mechanisms in Deep Learning (DL) models. Methods: We compared three attention-enhanced DL models, the brand-new InstaGATs, an LSTM with attention and a CNN with attention. We used these models to classify normal and abnormal (i.e., artifactual or pathological) EEG patterns. Results: We…
▽ More
Objective: To evaluate the impact on Electroencephalography (EEG) classification of different kinds of attention mechanisms in Deep Learning (DL) models. Methods: We compared three attention-enhanced DL models, the brand-new InstaGATs, an LSTM with attention and a CNN with attention. We used these models to classify normal and abnormal (i.e., artifactual or pathological) EEG patterns. Results: We achieved the state of the art in all classification problems, regardless the large variability of the datasets and the simple architecture of the attention-enhanced models. We could also prove that, depending on how the attention mechanism is applied and where the attention layer is located in the model, we can alternatively leverage the information contained in the time, frequency or space domain of the dataset. Conclusions: with this work, we shed light over the role of different attention mechanisms in the classification of normal and abnormal EEG patterns. Moreover, we discussed how they can exploit the intrinsic relationships in the temporal, frequency and spatial domains of our brain activity. Significance: Attention represents a promising strategy to evaluate the quality of the EEG information, and its relevance, in different real-world scenarios. Moreover, it can make it easier to parallelize the computation and, thus, to speed up the analysis of big electrophysiological (e.g., EEG) datasets.
△ Less
Submitted 2 December, 2020;
originally announced December 2020.
-
Complexity Issues of String to Graph Approximate Matching
Authors:
Riccardo Dondi,
Giancarlo Mauri,
Italo Zoppis
Abstract:
The problem of matching a query string to a directed graph, whose vertices are labeled by strings, has application in different fields, from data mining to computational biology. Several variants of the problem have been considered, depending on the fact that the match is exact or approximate and, in this latter case, which edit operations are considered and where are allowed. In this paper we pre…
▽ More
The problem of matching a query string to a directed graph, whose vertices are labeled by strings, has application in different fields, from data mining to computational biology. Several variants of the problem have been considered, depending on the fact that the match is exact or approximate and, in this latter case, which edit operations are considered and where are allowed. In this paper we present results on the complexity of the approximate matching problem, where edit operations are symbol substitutions and are allowed only on the graph labels or both on the graph labels and the query string. We introduce a variant of the problem that asks whether there exists a path in a graph that represents a query string with any number of edit operations and we show that is is NP-complete, even when labels have length one and in the case the alphabet is binary. Moreover, when it is parameterized by the length of the input string and graph labels have length one, we show that the problem is fixed-parameter tractable and it is unlikely to admit a polynomial kernel. The NP-completeness of this problem leads to the inapproximability (within any factor) of the approximate matching when edit operations are allowed only on the graph labels. Moreover, we show that the variants of approximate string matching to graph we consider are not fixed-parameter tractable, when the parameter is the number of edit operations, even for graphs that have distance one from a DAG. The reduction for this latter result allows us to prove the inapproximability of the variant where edit operations can be applied both on the query string and on graph labels.
△ Less
Submitted 7 January, 2020;
originally announced January 2020.
-
Top-k Overlapping Densest Subgraphs: Approximation and Complexity
Authors:
Riccardo Dondi,
Mohammad Mehdi Hosseinzadeh,
Giancarlo Mauri,
Italo Zoppis
Abstract:
A central problem in graph mining is finding dense subgraphs, with several applications in different fields, a notable example being identifying communities. While a lot of effort has been put on the problem of finding a single dense subgraph, only recently the focus has been shifted to the problem of finding a set of densest subgraphs. Some approaches aim at finding disjoint subgraphs, while in m…
▽ More
A central problem in graph mining is finding dense subgraphs, with several applications in different fields, a notable example being identifying communities. While a lot of effort has been put on the problem of finding a single dense subgraph, only recently the focus has been shifted to the problem of finding a set of densest subgraphs. Some approaches aim at finding disjoint subgraphs, while in many real-world networks communities are often overlapping. An approach introduced to find possible overlapping subgraphs is the Top-k Overlapping Densest Subgraphs problem. For a given integer k >= 1, the goal of this problem is to find a set of k densest subgraphs that may share some vertices. The objective function to be maximized takes into account both the density of the subgraphs and the distance between subgraphs in the solution.
The Top-k Overlapping Densest Subgraphs problem has been shown to admit a 1/10-factor approximation algorithm. Furthermore, the computational complexity of the problem has been left open. In this paper, we present contributions concerning the approximability and the computational complexity of the problem. For the approximability, we present approximation algorithms that improves the approximation factor to 1/2 , when k is bounded by the vertex set, and to 2/3 when k is a constant. For the computational complexity, we show that the problem is NP-hard even when k = 3.
△ Less
Submitted 30 January, 2019; v1 submitted 7 September, 2018;
originally announced September 2018.
-
Covering with Clubs: Complexity and Approximability
Authors:
Riccardo Dondi,
Giancarlo Mauri,
Florian Sikora,
Italo Zoppis
Abstract:
Finding cohesive subgraphs in a network is a well-known problem in graph theory. Several alternative formulations of cohesive subgraph have been proposed, a notable example being $s$-club, which is a subgraph where each vertex is at distance at most $s$ to the others. Here we consider the problem of covering a given graph with the minimum number of $s$-clubs. We study the computational and approxi…
▽ More
Finding cohesive subgraphs in a network is a well-known problem in graph theory. Several alternative formulations of cohesive subgraph have been proposed, a notable example being $s$-club, which is a subgraph where each vertex is at distance at most $s$ to the others. Here we consider the problem of covering a given graph with the minimum number of $s$-clubs. We study the computational and approximation complexity of this problem, when $s$ is equal to 2 or 3. First, we show that deciding if there exists a cover of a graph with three $2$-clubs is NP-complete, and that deciding if there exists a cover of a graph with two $3$-clubs is NP-complete. Then, we consider the approximation complexity of covering a graph with the minimum number of $2$-clubs and $3$-clubs. We show that, given a graph $G=(V,E)$ to be covered, covering $G$ with the minimum number of $2$-clubs is not approximable within factor $O(|V|^{1/2 -\varepsilon})$, for any $\varepsilon>0$, and covering $G$ with the minimum number of $3$-clubs is not approximable within factor $O(|V|^{1 -\varepsilon})$, for any $\varepsilon>0$. On the positive side, we give an approximation algorithm of factor $2|V|^{1/2}\log^{3/2} |V|$ for covering a graph with the minimum number of $2$-clubs.
△ Less
Submitted 4 June, 2018;
originally announced June 2018.
