Concept drift detectors are used in combination with learning systems to maintain a good accuracy rate on non-stationary data streams1. Financial time series are an instance of non-stationary data streams whose concept drifts (market phases) are so important to affect investment decisions worldwide. This paper studies how concept drift detectors behave when applied to financial time series and reports an experimentation on the SPY time series.

In recent years, the market of consumer products has experienced a growing interest in wireless power charging technologies and several embedded systems are now available for everyday use. Some of the largest electronics component manufacturers propose their own reference designs and system solutions, meeting international standard requirements of interoperability. Wireless Power Consortium (WPC) [1] is one of the leading wireless power standards that describes the interface of the so-called “Qi-standard” compliant devices. The Qi-standard defines the main features, which those certifiable devices must implement.

The paper describes a novel methodology to compare learning algorithms by exploiting their performance maps. A performance map enhances the comparison of a learner across learning contexts and it also provides insights in the distribution of a learners’ performances across its parameter space. Also some initial empirical findings are commented. In order to explain the novel comparison methodology, this study introduces the notions of learning context, performance map, and high performance function. These concepts are then applied to a variety of learning contexts to show how the methodology can be applied. Finally, we will use meta-optimization as an instrument to improve the efficiency of the parameter space search with respect to its complete enumeration. But, note that meta-optimization is neither an essential part of our methodology nor the focus of our study.

The paper discusses the notions of explainability and interpretability when using decision tree learning and agent based modeling to approximate financial time series. And how they related to the selected learning algorithm. As experimental context, the LFABS system for agent based modeling is used together with C4.5 for decision tree learning. The paper proposes the following definitions for interpretability: being able to make sense of a learning system output. And explainability: understanding how that output was generated. The study’s goal is achieved by comparing the different knowledge representations used by the two systems.

We will discuss the notions of explainability and interpretability when using agent based modeling to approximate market indexes. As working context we will use the L-FABS system [22, 28] where agent based modeling, whose parameters are learned by simulated annealing, is used to explain and predict financial time series like: SP500, DJIA, GLD, SLV, etc. We will assume the following definitions for interpretability: being able to make sense of system output, and explainability: understanding how that output was generated as in [18]. Novelty of the paper: the discussion of explainability and interpretability in agent based modelling as implemented in L-FABS. An empirical case study will be discussed. Please note that the goal of this paper is not to describe how L-FABS works.

The paper reports an empirical study done to statistically validate the preliminary findings obtained in previous research by the author on the small disjunct problem. Thus additional support to the working hypothesis that cooperative evolution (co-evolution) can be successfully applied in learning symbolic concepts and that co-evolution when carefully exploited can produce more robust classification rule (symbolic concepts) with higher statistical validity. In the paper we will compare the effect of applying a specific co-evolutive learning strategy with the results obtained by running a learning system without any coevolution. Thus we can measure the add-on effect produced by the coevolutive strategy. As learning systems we will use the system REGAL that combines distributed learning and genetic algorithms to find symbolic classifiers. As a future extension of this research, we note that the described co-evolutive strategy can be applied to other learning methods.

This discussion paper presents some general insights on the Italian technology transfer environment that the author has gained founding ProMarket 11, an Italian innovative startup, five years ago. During that period, ProMarket 11 has contacted several financial service institutions in Italy, all claiming to have internal R\&D centres or being startup friendly or being open innovation organizations, for setting up a pilot project to test the technology. The answers received are worth thinking about. The paper also proposes that no Italian Google (or other Italian multinational internet companies) exists because of the peculariaty of the Italian tech transfer environment. Finally the paper proposes how the Italian tech transfer environment could be radically changed by presenting an opposite view to that by Abravanel.

The paper proposes a computational adaptation of the principles underlying principal component analysis with agent based simulation in order to produce a novel modeling methodology for financial time series and financial markets. Goal of the proposed methodology is to find a reduced set of investor's models (agents) which is able to approximate or explain a target financial time series. As computational testbed for the study, the learning system L-FABS was chosen which combines simulated annealing with agent based simulation for approximating financial time series. Two experimental case studies showing the efficacy of the proposed methodology are reported.

Concept drift detectors allow learning systems to maintain good accuracy on non-stationary data streams. Financial time series are an instance of non-stationary data streams whose concept drifts (market phases) are so important to affect investment decisions worldwide. This paper studies how concept drift detectors behave when applied to financial time series. General results are: a) concept drift detectors usually improve the runtime over continuous learning, b) their computational cost is usually a fraction of the learning and prediction steps of even basic learners, c) it is important to study concept drift detectors in combination with the learning systems they will operate with, and d) concept drift detectors can be directly applied to the time series of raw financial data and not only to the model's accuracy one. Moreover, the study introduces three simple concept drift detectors, tailored to financial time series, and shows that two of them can be at least as effective as the most sophisticated ones from the state of the art when applied to financial time series. Currently submitted to Pattern Recognition

This work presents an extensive case study on modelling the DAX (Deutscher Aktienindex) index and United States Oil Fund (USO) exchange-traded fund (Etf) time series with the financial agent-based system learning financial agent-based simulator (L-FABS) that exploits simulated annealing as a learning method. The USO Etf time series is highly correlated with oil price behaviour, and the DAX index is based on the weighted and accumulated behaviour of the share prices of some of the largest companies traded on the Frankfurt Stock Exchange. These two time series are driven by completely different economic factors and thus provide two diverse empirical settings to evaluate the effectiveness of our methodology. Our experimentation shows that a relatively simple computational representation of real financial markets is effective in capturing the overall behaviour of the time series with varying approximation levels while the prediction target is moved into the future. The reported experimental investigation of L-FABS shows that it is robust notwithstanding the learning method used and the data sets exploited. L-FABS indeed produced a relatively low approximation error in several settings even when evaluated with respect to other modelling approaches, for example, 0.88% and 1.61% errors on average for 1 day ahead experiments in, respectively, DAX index and USO Etf.

Some employees suffer from burnout, and most bosses ignore this problem. Job burnout may hinder employees’ quality of life, personal accomplishment and satisfaction with life in general. It can also influence negatively the profits of the business or the organization, as the literature reveals. Mindfulness-based interventions have proven to be useful for ameliorating some aspects of burnout. By the same token, some agent-based simulator (ABSs) have been useful for predicting the influence of mindfulness programs on meditators in different aspects such as their emotions and their heart rate variability. In this context, the current work presents a novel ABS application that simulates the effects of mindfulness-based interventions on the job burnout subscales known as emotional exhaustion, depersonalization, personal accomplishment, exhaustion in general, and disengagement from work. This application allows users to define mindfulness programs without needing any computer-science technical knowledge and simulates its influence on a group of practitioners with certain features. The simulator has been tested by simulating two mindfulness-based interventions of two scenarios reported in the literature. The ABS received input from the pre-intervention burnout measures, and performed 1000 simulations for each scenario for avoiding bias from the nondeterministic behavior. The simulated outcomes referring to the post-intervention burnout measures were similar to the real ones. The mean differences, mean squared errors and mean absolute error were below 0.4% in the normalized values of all the burnout subscales reported in the two scenarios. The source code of this ABS is publicly available for guaranteeing reproducibility and allowing other researchers to extend it or reuse some of its components.

In this work we will report our initial investigation of how a neural network architecture could become an efficient tool to model Proportional-Integral-Derivative controller (PID controller). It is well known that neural networks are excellent function approximators, we will then be investigating if a recursive neural networks could be suitable to model and tune PID controllers thus could assist in determining the controller's proportional, integral, and the derivative gains. A preliminary evaluation is reported.

A computational approach combining machine learning (simulated annealing) and agent based simulation is shown to approximate financial time series. The agent based model allows to simulate the market conditions that produced the financial time series and simulated annealing optimize the parameters for the agent based model. The originality of our approach stays in the combination of financial market simulation with meta-learning of its parameters. The original contribution of the paper stays in discussing how the methodology can be applied under several meta-learning conditions and its experimentation on the real world SPDR Gold Trust (GLD) timeseries.

This paper reports a case study on modeling the SPDR Silver Trust (SLV) and Nasdaq Composite Index timeseries by using a financial agent based system using simulated annealing. We show here how adding financial information to the modeling system can significantly improve the modeling results. The learning system LFABS, previously developed by the author, will be used as a testbed for the empirical evaluation of the proposed methodology on the two case studies.

This paper presents a non-linear analysis of DC-shift induced by power supply noise in bandgap voltage references based on a diode connected PNP bipolar junction transistors (BJT) couple. The analysis is based on diode non-linear characteristic causing a variation of biasing currents involved in closed loop feedback. The observed effect is discussed, simulated, measured and compared with analytical derivation. A bandgap reference has been implemented in a CMOS 28 nm technology and used for successfully validating the proposed analysis.

This paper presents a 4.8 V tolerant circuit for reliably switching a startup load between a main power supply and a battery power supply. The circuit automatically switches the main power supply over to the battery in case the main line has been interrupted. The circuit includes a pair of back-to-back switch transistors for isolating the load from each power supply, a bias circuit for controlling the switch transistors, two independent current sources and two current subtraction units for deciding which supply to provide to the load. It consumes less than 1 μA per input and it can supply a startup circuitry up to 50 μA. The circuit has been implemented in a CMOS 28 nm technology, using only “low-voltage” devices and was successfully validated in the lab.

This paper presents an integrated bandgap reference circuit which is addressing low current consumption and a wide supply voltage range, using a current mode structure. Embedded in a sophisticated Power Management Unit (PMU) for a GNSS receiver, this bandgap reference has an output of 0.60 V and it can reach a temperature coefficient of 33 ppm/°C in the range from -40 °C to 125 °C. With a 1.4 V supply voltage, the power is only 3.5 μW and the PSRR is 57 dB at DC frequency. Occupying 0.125 mm2 chip area, this bandgap reference has been implemented in the CMOS 28 nm technology from Globalfoundries and successfully validated in the lab.

Wireless power transfer (WPT) systems are becoming ubiquitous with applications in powering medical implants and a range of portable consumer electronic devices such as smart phones and wearable devices. Wireless power transferring methods can be classified into two types: inductive and resonant. For the resonant type, wider-range power transfer is possible, and multiple devices with different power requirements can be charged at the same time. The Alliance for Wireless Power (A4WP) has chosen the 6.78MHz ISM band as the power-transfer frequency [1]. At 6.78MHz, the associated switching loss is an order of magnitude larger than that in a typical wireless receiver based on an inductive coupling, with a carrier frequency of around 200kHz. Besides the two fundamental aspects of switching frequency and power, there is a third important parameter, notably the higher input voltage range needed for the `loosely coupled' resonant type, which is 25V maximum for this work. Ref. [2] appears to be one of few works that can be entirely related to the current work, targeting the same application. However, it does not integrate the most critical parts of the receiver, such as the AC-DC rectifier. Other works in the same frequency range are either limited to low-power applications [3] or the AC-DC rectifier is a stand-alone chip [4-5].