In-memory computing technology is used extensively in artificial intelligence devices due to lower power consumption and fast calculation of matrix-based functions. The development of such a device and its integration in a system takes a significant amount of time and requires the use of a real-time emulation environment, where various system aspects are analyzed, microcode is tested, and applications are deployed, even before the real chip is available. In this work, we present the architecture, the software development tools, and experimental results of a distributed and expandable emulation system for rapid prototyping of integrated circuits based on in-memory computing technologies. Presented experimental results demonstrate the usefulness of the proposed emulator.

This paper presents an instruction-based coordination architecture for Field-Programmable Gate Array (FPGA)-based systems with multiple high-performance Processing Units (PUs) for accelerating Deep Neural Network (DNN) inference. This architecture enables programmable multi-PU synchronization through instruction controller units coupled with peer-to-peer instruction synchronization units, utilizing instruction types organized into load, compute, and store functional groups. A compilation framework is presented that transforms DNN models into executable instruction programs, enabling flexible partitioning of DNN models into topologically contiguous subgraphs mapped to available PUs. Multiple deployment strategies are supported, enabling pipeline parallelism among PUs and batch-level parallelism across different PU subsets, with runtime switching among them without FPGA reconfiguration. The proposed approach enables design space exploration, supporting dynamic trade-offs between single-batch and multi-batch performance. Experimental results on ResNet-50 demonstrate notable compute efficiency, up to 98%, and throughput efficiency gains, up to 2.7×, over prior works across different configurations.

Deep neural network (DNN) inference relies increasingly on specialized hardware for high computational efficiency. This work introduces a field-programmable gate array (FPGA)-based dynamically configurable accelerator featuring systolic arrays (SAs), high-bandwidth memory (HBM), and UltraRAMs. We present two processing unit (PU) configurations with different computing capabilities using the same interfaces and peripheral blocks. By instantiating multiple PUs and employing a heuristic weight transfer schedule, the architecture achieves notable throughput efficiency over prior works. Moreover, we outline how the architecture can be extended to emulate analog in-memory computing (AIMC) devices to aid next-generation heterogeneous AIMC chip designs and investigate device-level noise behavior. Overall, this brief presents a versatile DNN inference acceleration architecture adaptable to various models and future FPGA designs.

In this study, we explore the impact of COVID-19 pandemic on the default risk of loan portfolios of the Greek banking system, using cutting edge machine learning technologies, like deep learning. Our analysis is based on loan level monthly data, spanning a 42-month period, collected through the ECB AnaCredit database. Our dataset contains more than three million records, including both the pre- and post-pandemic periods. We develop a series of credit rating models implementing state of the art machine learning algorithms. Through an extensive validation process, we explore the best machine learning technique to build a behavioral credit scoring model and subsequently we investigate the estimated sensitivities of various features on predicting default risk. To select the best candidate model, we perform comparisons of the classification accuracy of the proposed methods, in 2-months out-of-time period. Our empirical results indicate that the Deep Neural Networks (DNN) have a superior predictive performance, signalling better generalization capacity against Random Forests, Extreme Gradient Boosting (XGBoost), and logistic regression. The proposed DNN model can accurately simulate the non-linearities caused by the pandemic outbreak on the evolution of default rates for Greek corporate customers. Under this multivariate setup we apply interpretability algorithms to isolate the impact of COVID-19 on the probability of default, controlling for the rest of the features of the DNN. Our results indicate that the impact of the pandemic peaks in the first year, and then it slowly decreases, though without reaching yet the pre COVID-19 levels. Furthermore, our empirical results also suggest different behavioral patterns between Stage 1 and Stage 2 loans, and that default rate sensitivities vary significantly across sectors. The current empirical work can facilitate a more in-depth analysis of AnaCredit database, by providing robust statistical tools for a more effective and responsive micro and macro supervision of credit risk.

The need to repeatedly shuttle around synaptic weight values from memory to processing units has been a key source of energy inefficiency associated with hardware implementation of artificial neural networks. Analog in-memory computing (AIMC) with spatially instantiated synaptic weights holds high promise to overcome this challenge, by performing matrix-vector multiplications (MVMs) directly within the network weights stored on a chip to execute an inference workload. However, to achieve end-to-end improvements in latency and energy consumption, AIMC must be combined with on-chip digital operations and communication to move towards configurations in which a full inference workload is realized entirely on-chip. Moreover, it is highly desirable to achieve high MVM and inference accuracy without application-wise re-tuning of the chip. Here, we present a multi-core AIMC chip designed and fabricated in 14-nm complementary metal-oxide-semiconductor (CMOS) technology with backend-integrated phase-change memory (PCM). The fully-integrated chip features 64 256x256 AIMC cores interconnected via an on-chip communication network. It also implements the digital activation functions and processing involved in ResNet convolutional neural networks and long short-term memory (LSTM) networks. We demonstrate near software-equivalent inference accuracy with ResNet and LSTM networks while implementing all the computations associated with the weight layers and the activation functions on-chip. The chip can achieve a maximal throughput of 63.1 TOPS at an energy efficiency of 9.76 TOPS/W for 8-bit input/output matrix-vector multiplications.

In this study, we propose a novel approach of nowcasting and forecasting the macroeconomic status of a country using deep learning techniques. We focus particularly on the US economy but the methodology can be applied also to other economies. Specifically US economy has suffered a severe recession from 2008 to 2010 which practically breaks out conventional econometrics model attempts. Deep learning has the advantage that it models all macro variables simultaneously taking into account all interdependencies among them and detecting non-linear patterns which cannot be easily addressed under a univariate modelling framework. Our empirical results indicate that the deep learning methods have a superior out-of-sample performance when compared to traditional econometric techniques such as Bayesian Model Averaging (BMA). Therefore our results provide a concise view of a more robust method for assessing sovereign risk which is a crucial component in investment and monetary decisions.

In this study, we address the topic of credit risk stemming from central governments from a technical point of view. First, we explore various econometric and machine learning techniques to build an enhanced sovereign rating system that effectively differentiates the risk of default among countries. Our empirical results indicate that the machine learning method of XGBOOST has a superior out-of-sample and out-of-time predictive performance. Then, we use the models developed to calibrate a sovereign rating system and provide useful insights into the set-up of a parsimonious early warning system. Our results provide a more concise view of the most robust method for classifying countries' default risk with significant regulatory implications, given that the efficient assessment of sovereign debt is crucial for effective proactive risk measurement.

We present a 256 $\times$ 256 in-memory compute (IMC) core designed and fabricated in 14-nm CMOS technology with backend-integrated multi-level phase change memory (PCM). It comprises 256 linearized current-controlled oscillator (CCO)-based A/D converters (ADCs) at a compact 4- $\mu \text{m}$ pitch and a local digital processing unit (LDPU) performing affine scaling and ReLU operations. A frequency-linearization technique for CCO is introduced, which increases the maximum CCO frequency beyond 3 GHz, while ensuring accurate on-chip matrix–vector multiplications (MVMs). Moreover, the design and functionality of the digital ADC calibration procedure is described in detail and the MVM accuracy is quantified. Finally, the measured classification accuracies of deep learning (DL) inference applications on the MNIST and CIFAR-10 datasets, when two IMC cores are employed, are presented. For a performance density of 1.59 TOPS/mm2, a measured energy efficiency of 10.5 TOPS/W, at a main clock frequency of 1 GHz, is achieved.

In the aftermath of the financial crisis, supervisory authorities have considerably altered the mode of operation of financial stress testing. Despite these efforts, significant concerns and extensive criticism have been raised by market participants regarding the considered unrealistic methodological assumptions and simplifications. Current stress testing methodologies attempt to simulate the risks underlying a financial institution’s balance sheet by using several satellite models. This renders their integration a really challenging task, leading to significant estimation errors. Moreover, advanced statistical techniques that could potentially capture the non-linear nature of adverse shocks are still ignored. This work aims to address these criticisms and shortcomings by proposing a novel approach based on recent advances in Deep Learning towards a principled method for Dynamic Balance Sheet Stress Testing. Experimental results on a newly collected financial/supervisory dataset, provide strong empirical evidence that our paradigm significantly outperforms traditional approaches; thus, it is capable of more accurately and efficiently simulating real world scenarios.

A core input in performing a regulatory stress test is the evolution of interest rates, as it affects the income generated from the assets’ side and the expenses from the liabilities’ side. In this work, we apply an autoregressive model with distributed lags (ADL) to quantify the pass through rates, that is, the degree and speed of incorporation of the changes of money market rates by banks into their customers deposit and loan rates. In doing so, for the liabilities’ side, we differentiate between open and term deposits, as well as between households and non-financial corporates. Our results indicate that for term deposits the long-term pass through rate is very high, exceeding 91% for non-financial corporate customers and 81% for households. For open deposits, the pass through rate dynamics appear less prevalent, amounting to 21% for non-financial corporate customers and 16% for households. When exploring the pass through rate dynamics in the assets’ side of the banks, we observe full long-term pass-through of money market rates, for mortgage and consumer loans. By contrast, the non-financial corporate loans rate is stickier and less reactive to money market rates changes, with long-term pass-through adjustment being approximately equal to 40%. Furthermore, our results provide evidence that the Greek sovereign spread movement has practically negligible pass through rate both for loan and deposit products. In particular, it hardly affects the pricing of new term deposits, with a pass through rate of around 5%. This finding can be attributed, among others factors, to the fact that the Greek sovereign credit spread has approached several times non-tradable territories, which makes it an insignificant variable in determining customer rates.

Central Bank speeches usually function as aggregators of internal quantitative and qualitative analysis of the institutions regarding the macro economy, the monetary policy and the health of the financial systems. Speeches usually function as a summary of the current status of a countries economic health, the undergoing trends and some future perspectives of the global economy. In this study departing from classical econometrics we employ natural language processing technologies in combination with machine learning techniques in order to filter out the most important signals in the corpus of speeches and translate into a sentiment index for forecasting the future financial markets behaviour. In our analysis, it is evident that central banker's expectations on economy tend to exhibit a predictive ability for financial markets turmoil. Using a combination of dictionaries which are either predefined or build based on historical speeches of the corpus we train an Extreme Gradient Boosting model that generates a sentiment index which signals turmoil with acceptable accuracy when passing a specific threshold.

We propose an innovative approach to model the probability of interlinkages in an interbank network with the use of Machine Learning techniques. More precisely we forecast the probability of a pair of banks entering into an interbank market borrower - lender relationship considering their financial characteristics and their past observed behavior. In this framework we examine a new method that employs machine learning in order to increase the accuracy of agnostic algorithms in reconstructing a financial network. The XGBOOST method is combined with both Maximum Entropy (MAXE) and Minimum Density (ANAN). The main contribution of this paper is that we enrich the information generally available for financial networks with variables that are available for the publication of banks financial statements (ensemble method). A set of agnostic models, i.e. models that the exposure allocation algorithm does not include prior information, are used as a benchmark to measure the additional benefit for applying machine learning in estimating prior network probabilities. By comparing the results between the agnostic algorithms and the ensemble method we see an increase in the accuracy and a decrease in the MAE of the financial networks on average. Our purpose is to depart from agnostic assumptions usually employed in interbank matrix allocation algorithms and take into account the financial features of the banks when assigning prior link probabilities. Our main finding is that machine learning algorithms outperforms the benchmark Logistic Regression model in interbank link forecasting and this is also reflected in the enhanced performance when overall network similarity measures are performed.

In this paper we apply text mining methodologies on a set of 10,000 Central Bank speeches to construct a financial dictionary, based on which we use Google Trends indices to measure people's interest in financial news. Particularly, we investigate the relationship between these indices and financial market turbulence leveraging on Deep Learning techniques, which are benchmarked against a variety of Machine Learning algorithms and traditional statistical techniques. Our main finding is that Google queries convey information able to predict future market turbulence in a short time period (one month), and that Deep Learning algorithms clearly outperform over benchmark techniques. Google Trends can provide useful input in the creation of crisis Early Warning Systems, as social data are more responsive compared to official financial indicators, which are usually available with a lag of several weeks or months. Thus, such an Early Warning System (EWS) that is continuously updated with current social data can be a valuable tool for policymakers, as it can immediately identify signs of whether a crisis is imminent or not.

We present a 256×256 in-memory compute (IMC) core designed and fabricated in 14nm CMOS with backend-integrated multi-level phase-change memory (PCM). It comprises 256 linearized current controlled oscillator (CCO)-based ADCs at a compact 4µm pitch and a local digital processing unit performing affine scaling and ReLU operations A novel frequency-linearization technique for CCOs is introduced, leading to accurate on-chip matrix-vector-multiply (MVM) when operating over 1 GHz. Measured classification accuracies on MNIST and CIFAR-10 datasets are presented when two cores are employed for deep learning (DL) inference The measured energy efficiency is 10.5 TOPS/W at a performance density of 1.59 TOPS/mm 2 .

The recent surge in Bitcoin price performance has attracted significant attention from both the market and academic researchers. This paper constitutes the first principled attempt to determine market risk own-funds requirements for Bitcoin. To this end, we examine price microstructure of the USD per bitcoin, and compare to other financial variables, as a proxy toward classifying Bitcoin into the appropriate risk-class. Using the outcomes of this analysis, we classify and quantify the entailed risk from a market risk minimum capital requirements perspective. To perform the prescribed analysis, we introduce a novel methodological paradigm, which adopts bleeding-edge concepts from the field of Data Science and Machine Learning.

Deep neural networks (DNNs) have revolutionized the field of artificial intelligence and have achieved unprecedented success in cognitive tasks such as image and speech recognition. Training of large DNNs, however, is computationally intensive and this has motivated the search for novel computing architectures targeting this application. A computational memory unit with nanoscale resistive memory devices organized in crossbar arrays could store the synaptic weights in their conductance states and perform the expensive weighted summations in place in a non-von Neumann manner. However, updating the conductance states in a reliable manner during the weight update process is a fundamental challenge that limits the training accuracy of such an implementation. Here, we propose a mixed-precision architecture that combines a computational memory unit performing the weighted summations and imprecise conductance updates with a digital processing unit that accumulates the weight updates in high precision. A combined hardware/software training experiment of a multilayer perceptron based on the proposed architecture using a phase-change memory (PCM) array achieves 97.73% test accuracy on the task of classifying handwritten digits (based on the MNIST dataset), within 0.6% of the software baseline. The architecture is further evaluated using accurate behavioral models of PCM on a wide class of networks, namely convolutional neural networks, long-short-term-memory networks, and generative-adversarial networks. Accuracies comparable to those of floating-point implementations are achieved without being constrained by the non-idealities associated with the PCM devices. A system-level study demonstrates 172 × improvement in energy efficiency of the architecture when used for training a multilayer perceptron compared with a dedicated fully digital 32-bit implementation.

Proactively monitoring and assessing the economic health of financial system has always been the cornerstone of supervisory authorities for supporting informed and timely decision making. Bank of Greece as the competent supervisory authority for the Greek banking system evaluates both the riskiness of banks on an individual level and the health of the financial system in total, from a macro prudential perspective. In accomplishing those targets, the Bank of Greece could make use –inter aliaof various statistical methods along with expert judgment. In this work, we employ a series of innovative modeling techniques in the prediction of individual bank insolvencies and generalized financial crises. Our empirical results indicate that innovative statistical techniques, i.e. Deep Learning and Machine Learning methodologies, have superior out of sample and out of time predictive performance in comparison to traditionally employed methods in finance, such as Logistic Regression, Classification Tress, and Linear Discriminant Analysis. In essence, we build an Early Warning System for bank insolvencies and another one for stock market crises, which could complement the assessments performed by micro-prudential and macroprudential authorities. In short, the holistic monitoring of the resilience of the financial system would steer decision making, via triggering the imposition of any necessary targeted corrective actions, leading vulnerable institutions back to viable business performance and the financial system back to balanced operation.

In-memory computing is an emerging non-von Neumann computing paradigm where certain computational tasks are performed in memory by exploiting the physical attributes of the memory devices. Memristive devices such as phase-change memory (PCM), where information is stored in terms of their conductance levels, are especially well suited for in-memory computing. In particular, memristive devices, when organized in a crossbar configuration can be used to perform matrix-vector multiply operations by exploiting Kirchhoff's circuit laws. To explore the feasibility of such in-memory computing cores in applications such as deep learning as well as for system-level architectural exploration, it is highly desirable to develop an accurate hardware emulator that captures the key physical attributes of the memristive devices. Here, we present one such emulator for PCM and experimentally validate it using measurements from a PCM prototype chip. Moreover, we present an application of the emulator for neural network inference where our emulator can capture the conductance evolution of approximately 400,000 PCM devices remarkably well.

The outbreak of the pandemic of COVID-19 in the beginning of 2020 had a significant impact both to the real economy and to the financial markets which exhibited high volatility movements of unprecedented scale. In this study we employ an AR(1)-GARCH(1,1) model with Student-t distributed standard errors in 8 jurisdictions to investigate the effects of the pandemic outbreak in the stock market returns, and based on that to assess the impact of a new massive wave of infections in the upcoming Winter. Our main finding is that there is strong negative relationship between new COVID-19 cases and stock market returns, whereas policy reactions had a beneficial impact only when combined with the effectiveness of measures taken by each country to curtail the pandemic. Finally, even though the perception about the virus has changed significantly, as fear has given ground to the hope for treatment, an abrupt massive wave of new infections could provoke exacerbated financial turmoil in 2021.