Forecasting Financial Time Series Using Robust Deep Adaptive Input Normalization

Deep Learning provided powerful tools for forecasting financial time series data. However, despite the success of these approaches on many challenging financial forecasting tasks, it is not always straightforward to employ DL-based approaches for highly volatile and non-stationary time financial series. To this end, in this paper, an adaptive input normalization layer that can learn to identify the distribution from which the input data were generated and then apply the most appropriate normalization scheme is proposed. This allows for promptly adapting the input to the subsequent DL model, which can be especially important, given recent findings that hint at the existence of critical learning periods in neural networks. Furthermore, the proposed method operates on a sliding window over the time series allowing for overcoming non-stationary issues that often arise. It is worth noting that the main difference with existing approaches is that the proposed method does not just learn to perform static normalization, e.g., using a fixed set of parameters, but instead it adaptively calculates the most appropriate normalization parameters, significantly improving the robustness of the proposed approach when distribution shifts occur. The effectiveness of the proposed formulation is verified using extensive experiments on three challenging financial time-series datasets.

     

A decision-theoretic approach for model interpretability in Bayesian framework

Machine Learning - A salient approach to interpretable machine learning is to restrict modeling to simple models. In the Bayesian framework, this can be pursued by restricting the model structure...     

Clustering Structure Analysis in Time-Series Data With Density-Based Clusterability Measure

Clustering is used to gain an intuition of the structures in the data. Most of the current clustering algorithms produce a clustering structure even on data that do not possess such structure. In these cases, the algorithms force a structure in the data instead of discovering one. To avoid false structures in the relations of data, a novel clusterability assessment method called density-based clusterability measure is proposed in this paper. It measures the prominence of clustering structure in the data to evaluate whether a cluster analysis could produce a meaningful insight to the relationships in the data. This is especially useful in time-series data since visualizing the structure in time-series data is hard. The performance of the clusterability measure is evaluated against several synthetic data sets and time-series data sets, which illustrate that the density-based clusterability measure can successfully indicate clustering structure of time-series data.      

Damping properties of austenitic stainless steels containing strain-induced martensite

Damping properties of two austenitic stainless steel grades, EN 1.4318 and EN 1.4301, were investigated. The test materials were cold rolled to different reductions and damping capacity was measured as a function of temperature with an internal friction method. Microstructures of the test materials were studied by means of X-ray diffraction (XRD) and magnetic measurements. The results showed that damping capacity of the studied materials depended on the amounts of strain-induced ε- and α′-martensite phases. At temperatures around 0 °C, highest damping capacity was achieved with cold rolling reduction of 10 to 15 pct. This behavior is related to the existence of ε-martensite and stacking faults. Internal friction peak due to α′-martensite phase was present at the temperature of 130 °C. Strain aging heat treatment at 200 °C for 20 minutes decreased the damping capacity in the entire studied temperature range.     

Path sensitisation and gate sizing approach to low poweroptimisation

A powerful circuit optimisation solution for high-performance and low-power design is presented. The proposed method combines path sensitisation and gate resizing approaches to reduce the power dissipation under a given timing constraint. The algorithm is tested on ISCAS-85 benchmark circuits, and up to 30%, power reduction is achieved     

An open-source population balance modeling framework for the simulation of polydisperse multiphase flows

Polydispersity is a challenging feature of many industrial and environmental multiphase flows, influencing all related transfer and transport processes. Besides their size, the fluid or solid particles may be distributed with respect to other properties such as their velocity or shape. Here, a population balance model based on the method of classes is combined with a multifluid solver within the open source computational fluid dynamics library OpenFOAM. The model allows for tracking the evolution of one or more size-conditioned secondary properties. It is applied to two different problems, the first being bubbly flow of air and water in a vertical pipe, where considering the velocity as a secondary property allows to resolve the size-dependent radial segregation. The second application is the gas phase synthesis of titania powder, where non-spherical particle aggregates appear whose shape is modeled through a collision diameter, leading to an improved prediction of the size distribution.     

An adaptive fuzzy control system to maximize rough turning productivity and avoid the onset of instability

This paper presents a new method to improve cutting efficiency for steel rough turning. To date, most efforts aimed at improving productivity during cutting operations have concentrated on optimizing material handling to and from the machinery. Here, the focus is on improving the efficiency of the turning operation itself. The approach is to control feed rate to raise machine power to a maximum safe level while avoiding the onset of cutting instability. The measure of machine power comes directly from the spindle motor and is held below the cutting machine??s power capacity. Detecting the onset of instability relies on interpreting data that come from installed instrumentation. A fuzzy inference system processes the inputs and makes the final control decisions. The prototype system was tuned using data collected in a variety of cutting situations. Subsequent testing of the tuned control system showed that it was capable of successfully maximizing power usage while still avoiding the onset of instability.