No Arabic abstract
Data stream mining extracts information from large quantities of data flowing fast and continuously (data streams). They are usually affected by changes in the data distribution, giving rise to a phenomenon referred to as concept drift. Thus, learning models must detect and adapt to such changes, so as to exhibit a good predictive performance after a drift has occurred. In this regard, the development of effective drift detection algorithms becomes a key factor in data stream mining. In this work we propose CU RIE, a drift detector relying on cellular automata. Specifically, in CU RIE the distribution of the data stream is represented in the grid of a cellular automata, whose neighborhood rule can then be utilized to detect possible distribution changes over the stream. Computer simulations are presented and discussed to show that CU RIE, when hybridized with other base learners, renders a competitive behavior in terms of detection metrics and classification accuracy. CU RIE is compared with well-established drift detectors over synthetic datasets with varying drift characteristics.
As next-generation networks materialize, increasing levels of intelligence are required. Federated Learning has been identified as a key enabling technology of intelligent and distributed networks; however, it is prone to concept drift as with any machine learning application. Concept drift directly affects the models performance and can result in severe consequences considering the critical and emergency services provided by modern networks. To mitigate the adverse effects of drift, this paper proposes a concept drift detection system leveraging the federated learning updates provided at each iteration of the federated training process. Using dimensionality reduction and clustering techniques, a framework that isolates the systems drifted nodes is presented through experiments using an Intelligent Transportation System as a use case. The presented work demonstrates that the proposed framework is able to detect drifted nodes in a variety of non-iid scenarios at different stages of drift and different levels of system exposure.
Gliders in one-dimensional cellular automata are compact groups of non-quiescent and non-ether patterns (ether represents a periodic background) translating along automaton lattice. They are cellular-automaton analogous of localizations or quasi-local collective excitations travelling in a spatially extended non-linear medium. They can be considered as binary strings or symbols travelling along a one-dimensional ring, interacting with each other and changing their states, or symbolic values, as a result of interactions. We analyse what types of interaction occur between gliders travelling on a cellular automaton `cyclotron and build a catalog of the most common reactions. We demonstrate that collisions between gliders emulate the basic types of interaction that occur between localizations in non-linear media: fusion, elastic collision, and soliton-like collision. Computational outcomes of a swarm of gliders circling on a one-dimensional torus are analysed via implementation of cyclic tag systems.
Signals are a classical tool used in cellular automata constructions that proved to be useful for language recognition or firing-squad synchronisation. Particles and collisions formalize this idea one step further, describing regular nets of colliding signals. In the present paper, we investigate the use of particles and collisions for constructions involving an infinite number of interacting particles. We obtain a high-level construction for a new smallest intrinsically universal cellular automaton with 4 states.
A simulation approach to the stochastic growth of bacterial towers is presented, in which a non-uniform and finite nutrient supply essentially determines the emerging structure through elementary chemotaxis. The method is based on cellular automata and we use simple, microscopic, local rules for bacterial division in nutrient-rich surroundings. Stochastic nutrient diffusion, while not crucial to the dynamics of the total population, is influential in determining the porosity of the bacterial tower and the roughness of its surface. As the bacteria run out of food, we observe an exponentially rapid saturation to a carrying capacity distribution, similar in many respects to that found in a recently proposed phenomenological hierarchical population model, which uses heuristic parameters and macroscopic rules. Complementary to that phenomenological model, the simulation aims at giving more microscopic insight into the possible mechanisms for one of the recently much studied bacterial morphotypes, known as towering biofilm, observed experimentally using confocal laser microscopy. A simulation suggesting a mechanism for biofilm resistance to antibiotics is also shown.
In user targeting automation systems, concept drift in input data is one of the main challenges. It deteriorates model performance on new data over time. Previous research on concept drift mostly proposed model retraining after observing performance decreases. However, this approach is suboptimal because the system fixes the problem only after suffering from poor performance on new data. Here, we introduce an adversarial validation approach to concept drift problems in user targeting automation systems. With our approach, the system detects concept drift in new data before making inference, trains a model, and produces predictions adapted to the new data. We show that our approach addresses concept drift effectively with the AutoML3 Lifelong Machine Learning challenge data as well as in Ubers internal user targeting automation system, MaLTA.