UH Biocomputation Group - Yi Sunhttp://biocomputation.herts.ac.uk/2022-05-26T15:31:48+01:00On Projection Methods for Functional Time Series Forecasting2022-05-26T15:31:48+01:002022-05-26T15:31:48+01:00Yi Suntag:biocomputation.herts.ac.uk,2022-05-26:/2022/05/26/on-projection-methods-for-functional-time-series-forecasting.html<p class="first last">Yi Sun's Journal Club session where she will talk about a paper "On Projection Methods for Functional Time Series Forecasting"</p>
<p>This week on Journal Club session Yi Sun will talk about a paper "On Projection Methods for Functional Time Series Forecasting".</p>
<hr class="docutils" />
<p>Two nonparametric methods are presented for forecasting functional time series
(FTS). The FTS we observe is a curve at a discrete-time point. We address both
one-step-ahead forecasting and dynamic updating. Dynamic updating is a forward
prediction of the unobserved segment of the most recent curve. Among the two
proposed methods, the first one is a straightforward adaptation to FTS of the
k-nearest neighbors methods for univariate time series forecasting. The second
one is based on a selection of curves, termed the curve envelope, that aims to
be representative in shape and magnitude of the most recent functional
observation, either a whole curve or the observed part of a partially observed
curve. In a similar fashion to k-nearest neighbors and other projection methods
successfully used for time series forecasting, we "project" the k-nearest
neighbors and the curves in the envelope for forecasting. In doing so, we keep
track of the next period evolution of the curves. The methods are applied to
simulated data, daily electricity demand, and NOx emissions and provide
competitive results with and often superior to several benchmark predictions.
The approach offers a model-free alternative to statistical methods based on
FTS modeling to study the cyclic or seasonal behavior of many FTS.</p>
<div class="line-block">
<div class="line"><br /></div>
</div>
<p>Papers:</p>
<ul class="simple">
<li>A. Elías, R. Jiménez, H. Shang, <a class="reference external" href="https://doi.org/10.1016/j.jmva.2021.104890">"On Projection Methods for Functional Time Series Forecasting"</a>, 2022, Journal of Multivariate Analysis, 189, 104890</li>
</ul>
<p><strong>Date:</strong> 2022/05/27 <br />
<strong>Time:</strong> 14:00 <br />
<strong>Location</strong>: online</p>
Clustering Functional Data into Groups by Using Projections2021-07-14T10:51:50+01:002021-07-14T10:51:50+01:00Yi Suntag:biocomputation.herts.ac.uk,2021-07-14:/2021/07/14/clustering-functional-data-into-groups-by-using-projections.html<p class="first last">Yi Sun's Journal Club session where she will talk about a paper "Clustering Functional Data into Groups by Using Projections"</p>
<p>This week on Journal Club session Yi Sun will talk about a paper "Clustering Functional Data into Groups by Using Projections".</p>
<hr class="docutils" />
<p>We show that, in the functional data context, by appropriately exploiting the
functional nature of the data, it is possible to cluster the observations
asymptotically perfectly. We demonstrate that this level of performance can
sometimes be achieved by the k -means algorithm as long as the data are
projected on a carefully chosen finite dimensional space. In general, the
notion of an ideal cluster is not clearly defined. We derive our results in the
setting where the data come from two populations whose distributions differ at
least in terms of means, and where an ideal cluster corresponds to one of these
two populations. We propose an iterative algorithm to choose the projection
functions in a way that optimizes clustering performance, where, to avoid
peculiar solutions, we use a weighted least squares criterion. We apply our
iterative clustering procedure on simulated and real data, where we show that
it works well.</p>
<div class="line-block">
<div class="line"><br /></div>
</div>
<p>Papers:</p>
<ul class="simple">
<li>A. Delaigle, P. Hall, T. Pham, <a class="reference external" href="https://doi.org/10.1111/rssb.12310">"Clustering Functional Data into Groups by Using Projections"</a>, 2019, Journal of the Royal Statistical Society: Series B (Statistical Methodology), 81, 271--304</li>
</ul>
<p><strong>Date:</strong> 2021/07/16 <br />
<strong>Time:</strong> 14:00 <br />
<strong>Location</strong>: online</p>
Asynchronism-based principal component analysis for time series data mining2018-10-10T14:33:32+01:002018-10-10T14:33:32+01:00Yi Suntag:biocomputation.herts.ac.uk,2018-10-10:/2018/10/10/asynchronism-based-principal-component-analysis-for-time-series-data-mining.html<p class="first last">Yi Sun's journal club session, where she will present the paper "<a class="reference external" href="https://www.sciencedirect.com/science/article/pii/S095741741300835X">Asynchronism-based principal component analysis for time series data mining (Hailin Li, 2014)</a>".</p>
<p>Yi Sun's journal club session, where she will present the paper "<a class="reference external" href="https://www.sciencedirect.com/science/article/pii/S095741741300835X">Asynchronism-based principal component analysis for time series data mining (Hailin Li, 2014)</a>".</p>
<hr class="docutils" />
<p>Principal component analysis (PCA) is often applied to dimensionality reduction for time series data mining. However, the principle of PCA is based on the synchronous covariance, which is not very effective in some cases. In this paper, an asynchronism-based principal component analysis (APCA) is proposed to reduce the dimensionality of univariate time series. In the process of APCA, an asynchronous method based on dynamic time warping (DTW) is developed to obtain the interpolated time series which derive from the original ones. The correlation coefficient or covariance between the interpolated time series represents the correlation between the original ones. In this way, a novel and valid principal component analysis based on the asynchronous covariance is achieved to reduce the dimensionality. The results of several experiments demonstrate that the proposed approach APCA outperforms PCA for dimensionality reduction in the field of time series data mining.</p>
<p><strong>Date:</strong> 12/10/2018 <br />
<strong>Time:</strong> 16:00 <br />
<strong>Location</strong>: LB252</p>
Singular Spectrum Analysis - A very short introduction2018-03-21T17:10:23+00:002018-03-21T17:10:23+00:00Yi Suntag:biocomputation.herts.ac.uk,2018-03-21:/2018/03/21/singular-spectrum-analysis-a-very-short-introduction-.html<p class="first last">Yi Sun's journal club session on Singular Spectrum Analysis</p>
<p>Singular Spectrum Analysis (SSA) can be used for smoothing time series and extracting the trend and oscillatory components. In this talk, Yi will introduce the basic idea of SSA and explain how it works by giving a simple example.</p>
<p><strong>Date:</strong> 23/03/2018 <br />
<strong>Time:</strong> 16:00 <br />
<strong>Location</strong>: LB252</p>
Feature Selection Modelling for Percutaneous Absorption across Synthetic Membranes2017-12-12T15:04:06+00:002017-12-12T15:04:06+00:00Yi Suntag:biocomputation.herts.ac.uk,2017-12-12:/2017/12/12/feature-selection-modelling-for-percutaneous-absorption-across-synthetic-membranes.html<p class="first last">Yi will be discussing feature selection modelling for percutaneous absorption across Synthetic membranes</p>
<p>Yi will be discussing feature selection modelling for percutaneous absorption across Synthetic membranes</p>
<p>Abstract is below:</p>
<hr class="docutils" />
<p>Predicting the rate of percutaneous absorption across mammalian and artificial membranes is a complex problem.
Different machine learning models have been used in previous studies, and results show that Gaussian processes provided the best result, based on a range of statistical measures.
In the current study, a dataset of synthetic (Polydimethylsiloxane, PDMS) membranes, containing so many descriptors, is considered.
One of the main purposes of the study is to use feature selection methods to select a set of molecular properties that exert the most important influence on percutaneous absorption across PDMS membranes, in the hope that this will better inform studies on human skin.</p>
<p><strong>Date:</strong> 15/12/2017 <br />
<strong>Time:</strong> 16:00 <br />
<strong>Location</strong>: LB252</p>
Dynamic functional principal components2016-11-29T10:43:11+00:002016-11-29T10:43:11+00:00Yi Suntag:biocomputation.herts.ac.uk,2016-11-29:/2016/11/29/dynamic-functional-principal-components.html<p class="first last">Yi Sun's journal club session on dynamic functional principal components.</p>
<p>In this presentation, I shall talk about S. Hofmann and L. Kidzinski's research paper, entitled Dynamic functional principal components. The paper has been published in Journal of the Royal Statistical Society: Series B (Statistical Methodology), Volume 77, Issue 2, March 2015, Pages 319–348: <a class="reference external" href="http://onlinelibrary.wiley.com/doi/10.1111/rssb.12076/full">http://onlinelibrary.wiley.com/doi/10.1111/rssb.12076/full</a></p>
<p>I shall introduce some basic background knowledge rather than discussing mathematical equations in the paper. Especially, I shall talk about the idea of Functional Principal Components, of Dynamic Principal Components, and how these two methods are combined together to be used in analysing time series data in this paper.</p>
<p><strong>Date:</strong> 02/12/2016 <br />
<strong>Time:</strong> 16:00 <br />
<strong>Location</strong>: LB252</p>
Development of computational models for characterizing small ice particles based on their two-dimensional light scattering patterns2016-01-19T17:46:06+00:002016-01-19T17:46:06+00:00Yi Suntag:biocomputation.herts.ac.uk,2016-01-19:/2016/01/19/development-of-computational-models-for-characterizing-small-ice-particles-based-on-their-two-dimensional-light-scattering-patterns.html<p class="first last">Yi Sun's journal club session on the development of computational models for characterizing small ice particles based on their two-dimensional light scattering patterns.</p>
<p>Clouds and their constituents (mostly small particles) have relationships with climate change. Better understanding and characterization of cloud particles is essential for the creation of better climate models for better understanding of climate and its dynamics/changes.</p>
<p>The widely used cloud particle imager has resolution limitations. Two-dimensional light scattering patterns (2DLS patterns) of particles is an increasingly more promising approach for characterising cloud particles.</p>
<p>Existing methods for characterizing cloud particles based on their 2DLS patterns perform poorly especially for patterns with too little speckles or with numerous fringes.</p>
<p>In this talk, I will focus on the background of the project, and on how we used Zernike moments as features which are insensitive to image rotation to obtain ice particle size and aspect ratio.</p>
<p><strong>Date:</strong> 22/01/2016 <br />
<strong>Time:</strong> 16:00 <br />
<strong>Location</strong>: LB252</p>