Edinburgh DataShare
Not a member yet
7718 research outputs found
Sort by
Achieving Robust Channel Estimation Neural Networks by Designed Training Data
No full textCode for Luan, Dianxin, and John Thompson. "Achieving Robust Channel Estimation Neural Networks by Designed Training Data.", Accepted for Publication in IEEE Transactions on Cognitive Communications and Networking (2025).
This dataset is uploaded from this Github page, maintained by Dianxin Luan:
https://github.com/dianixn/Achieving-Robust-Channel-Estimation-Neural-Networks-by-Designed-Training-Data
Paper Abstract: Channel estimation is crucial in wireless communications. However, in many papers neural networks are frequently tested by training and testing on one example channel or similar channels. This is because data-driven methods often degrade on new data which they are not trained on, as they cannot extrapolate their training knowledge. This is despite the fact physical channels are often assumed to be time-variant. However, due to the low latency requirements and limited computing resources, neural networks may not have enough time and computing resources to execute online training to fine-tune the parameters. This motivates us to design offline-trained neural networks that can perform robustly over wireless channels, but without any actual channel information being known at design time. In this paper, we propose design criteria to generate synthetic training datasets for neural networks, which guarantee that after training the resulting networks achieve a certain mean squared error (MSE) on new and previously unseen channels. Therefore, trained neural networks require no prior channel information or parameters update for real-world implementations. Based on the proposed design criteria, we further propose a benchmark design which ensures intelligent operation for different channel profiles. To demonstrate general applicability, we use neural networks with different levels of complexity to show that the generalization achieved appears to be independent of neural network architecture. From simulations, neural networks achieve robust generalization to wireless channels with both fixed channel profiles and variable delay spreads.
Funding Acknowledgement:
This research is supported by EPSRC projects EP/X04047X/2 and EP/Y037243/1 (TITAN Extension project).COPY OF README.TXT FILE IN THE ZIP FILE:
-----------------------------------------------------------------
# Achieving Robust Channel Estimation Neural Networks by Designed Training Data
Code for Luan, Dianxin, and John Thompson. "Achieving Robust Channel Estimation Neural Networks by Designed Training Data." IEEE Transactions on Cognitive Communications and Networking (2025).
Abstract:
Channel estimation is crucial in wireless communications. However, in many papers neural networks are frequently tested by training and testing on one example channel or similar channels. This is because data-driven methods often degrade on new data which they are not trained on, as they cannot extrapolate their training knowledge. This is despite the fact physical channels are often assumed to be time-variant. However, due to the low latency requirements and limited computing resources, neural networks may not have enough time and computing resources to execute online training to fine-tune the parameters. This motivates us to design offline-trained neural networks that can perform robustly over wireless channels, but without any actual channel information being known at design time. In this paper, we propose design criteria to generate synthetic training datasets for neural networks, which guarantee that after training the resulting networks achieve a certain mean squared error (MSE) on new and previously unseen channels. Therefore, trained neural networks require no prior channel information or parameters update for real-world implementations. Based on the proposed design criteria, we further propose a benchmark design which ensures intelligent operation for different channel profiles. To demonstrate general applicability, we use neural networks with different levels of complexity to show that the generalization achieved appears to be independent of neural network architecture. From simulations, neural networks achieve robust generalization to wireless channels with both fixed channel profiles and variable delay spreads.
%%%
Run Demonstration_of_H_Rayleigh_Propogation_Channel, Demonstration_of_H_Rayleigh_Propogation_Channel_Alternative and Demonstration_of_H_Rayleigh_Propogation_Channel_batch files to test for sub-6 band results.
Run Demonstration_of_H files to test for Millimeter-wave band results.
Run Demonstration_of_H_Appendix to get the first three figures of Simulation Section.
%% File +Training has
ResNN_pilot_regression to train the InterpolateNet and SimpleNet for default pilot pattern.
ResNN_pilot_regression_Alternative to train the InterpolateNet and SimpleNet for alternative pilot pattern.
Training_hybrid_offline to train Channelformer.
Training_hybrid_offline_Alternative to train Channelformer for alternative pilot pattern.
%% File +parameter has
parameters contains the system parameters for generating the training data and testing on the default pilot pattern on sub-6 band.
parameters_alternative contains the system parameters for generating the training data and testing on the alternative pilot pattern on sub-6 band.
parameters_hybrid contains the hyperparameters for Channelformer.
parameters_Millimeterwave contains the hyperparameters for CDL/TDL channels operating on Millimeter-wave band (39GHz).
%% File +Channel contains
Propagation_Channel_Model is a LTEfading channel developed by MATLAB specificed in https://uk.mathworks.com/help/lte/ref/ltefadingchannel.html.
CDL_Channel contains 3GPP TS38.901 CDL channel - nrCDLChannel object.
TDL_Channel contains 3GPP TS38.901 TDL channel - nrTDLChannel object.
%% File +CSI has
LS - It is the implementation of the LS method and the time interpolation method is bilinear method.
MMSE - It is the linear MMSE method and the time interpolation method is bilinear method.
%% File +Data_Generation contains
Data_Generation - used to generate the training data for online Channelformer offline
Data_Generation_Online - used to generate the training data for online training
Data_generation_offline_version - used to generate the training data for offline Channelformer and HA02.
Data_Generation_Residual - used to generate the training data for InterpolateNet and ReEsNet
Data_Generation_Transformer - used to generate the training data for TR method.
%% File +OFDM contains
OFDM_Receiver - OFDM receiver
OFDM_Transmitter - OFDM transmitter
Pilot_extract - extract the pilot
Pilot_Insert - insert the pilot
QPSK_Modualtor - generate QPSK symbols
QPSK_Demodulator - decode the received QPSK signals
%% File +Performance_plot contains
Plot_Alternative - plot function.
Plot_Appendix - plot function.
Plot_BER - plot function.
Plot_HA02 - plot function.
Plot_InterpolateNet - plot function.
Plot_N_f - plot function.
Plot_SimpleNet - plot function.
Plot_batch - plot function.
Plot_generalization - plot function.
%% File Residual_NN contains
Interpolation_ResNet - Untrained InterpolateNet (WSA paper)
SimpleNet - 882 parameters neural networks
%% File +transformer contains Channelformer code
model - system model for Channelformer
+HA03 - The encoder and decoder architecture of Channelformer
Encoder_block - the encoder of Channelformer
Decoder_block - the decoder of Channelformer
+layer contains the layer modules for attanetion mechanism and residual convolutional neural network
normalization - layer normalization
FC1 - fully-connected layer
gelu - Activation function of GeLu
multiheadAttention - multihead attention module, which calcualte the attention from Q, K and V
attention - main control unite of the multiohead attention module, designed by tranformer encoder
FeedforwardNN - feedforward neural network designed by tranformer encoder
%%% Comments
Run with MATLAB 2023B, with fully-installed deep learning toolbox because it requires customized training.
%%% Acknowledgement
This research is supported by UK Engineering and Physical Sciences Research Council (EPSRC) projects EP/X04047X/2 and EP/Y037243/1 for the TITAN Telecoms Hub
Care Home Stakeholder Perspectives on Data Sharing and Linkage: Search and Data Extraction
No full textThe data deposited relate to a rapid scoping review that I conducted under the Advanced Care Research Centre as part of my Individual Project in Ageing and Care. The following is a draft abstract for a pending publication of this review: "Since the COVID-19 pandemic highlighted inadequacies in the collection, access, and use of health and social care data across the United Kingdom (UK), the topics of ‘data sharing’ and ‘data linkage’ have gained greater importance. Yet, the sharing and linkage of social care data, especially that from care homes, remains underexplored. To address the limited understanding of care home stakeholders’ perspectives on data sharing and linkage, a rapid scoping review was conducted to map the extent and range of existing evidence on the topic and identify gaps in the literature. Searches were conducted in PubMed, EMBASE, CINAHL Plus, Scopus, Web of Science, and the Cochrane Database of Systematic Reviews to identify studies on stakeholders’ needs, views, and experiences of data sharing or linkage in the UK care home sector. After removing 322 duplicates, 543 studies were screened. The remaining 11 studies identified key barriers, facilitators, and opportunities to improve data sharing and linkage within the UK care home sector. The findings displayed that, while there are similarities between health and social care data sharing and linkage, further research is required to provide nuanced insights into the specific needs of the care home sector
“What, How, Why and Growing”: Sports Practitioners' Views on Teaching Methods Used with Young Learners
No full textThis investigation aimed to understand the factors influencing educators' opinions, preferences, perceptions, enhancements, and sources of information on the best ways to teach specific sports skills to children and young people. Using personal contacts in the industry and subsequent snowball sampling, survey links were sent to educators in Kuwait and the UK to collect data, and 155 participated. Participants generally agreed on the goals of teaching and the significance of achieving these goals. Notably, however, participants differed regarding the specific teaching methods that should be used to achieve these goals. Furthermore, participants supported the idea of using a variety teaching methods in the physical activities rather than only one. Lastly, it was found that educators use different teaching methods and face challenges in their teaching roles. Educators should be equipped to use a hybrid approach incorporating different methods depending on contextual factors. This will ensure that all learners are catered for and improve education quality.Data Description File
Survey Raw Datase
Explosive spore dispersal of Selaginella kraussiana
No full textVideos showing the explosive spore dispersal of the reproductive structures (strobili) in the lycophyte Selaginella kraussiana. In several instances in this dataset, microsporangia under desiccating conditions can be seen to firstly dehisce, exposing the microspores. This is then followed by the strobili mechanically ejecting the entire sporangium, resulting in an explosive projectile spore dispersal mechanism. This type of spore dispersal in lycophytes has been noted previously in the literature (Koller & Scheckler, 1986 https://doi.org/10.2307/2444062), but to our knowledge this is the first time that it has been recorded on video. Here, we deposit the full unedited videos. Videos are taken using a VHX-7000 Keyence light microscope
A subset of voltage-gated potassium channel data from Channelpedia.epfl used in the paper "One model to rule them all: unification of voltage-gated potassium channel models via deep non-linear mixed effects modelling"
No full textRaw cell-level .nwb data files used in the publication "One model to rule them all: unification of voltage-gated potassium channel models via deep non-linear mixed effects modelling", which can be processed using the code in the github repository https://github.com/dom-linkevicius/SciMLHHModels.jl to reproduce the results of the paper.For the description of the data see "A Kinetic Map of the Homomeric Voltage-Gated Potassium Channel (Kv) Family" https://doi.org/10.3389/fncel.2019.0035
Hong Kong November 2023 - images of new and existing estates
No full text### DOCOMOMO INTERNATIONAL MASS HOUSING ARCHIVE ### The provision of healthy modern housing for all was one of the foremost ideals of the Modern Movement, and inspired a vast wave of planning and building across the world during the 20th century. In the last quarter of the century, even as the foundational programmes of Europe and America lost their impetus, the baton was passed on to other countries, especially in eastern Asia, where the narrative of Modern mass housing was reinvigorated for the next century - a unique example of a key Modernist project that actually continues and thrives today, and which thus forms a principal focus of interest for DOCOMOMO – the leading international organisation promoting the documentation and conservation of buildings, sites and neighbourhoods of the Modern Movement. As heritage, the built legacies of this diverse and multi-generational adventure are almost always too controversial to qualify for conservation strategies. Instead, therefore, recording and inventorisation must dominate the heritage interest in this field. In the recognition of that fact, DOCOMOMO’s International Specialist Committee on Urbanism and Landscape, in partnership with the Scottish Centre for Conservation Studies at the University of Edinburgh, has launched the International Mass Housing Archive, whose aim is to provide an open-access library of images of significant housing projects in each working-group territory, free of copyright restrictions. These files may be copied, edited and shared on condition the appropriate citation is used, as per the terms of the attached Creative Commons Attribution licence. ### Structure ### The International Mass Housing Archive is subdivided under geographical headings corresponding to the constituent working groups of DOCOMOMO, and the individual housing projects are searchable under city and project name. Initially, the Image Archive will be managed and augmented centrally by DOCOMOMO and the SCCS, in partnership with University of Edinburgh Information Services, commencing with pilot city surveys sourced from our own photographic records in the first instance. The archive is related to several existing mass housing documentation initiatives. These include one concerning Britain, namely the online version of the 1994 book, Tower Block: http://towerblock.org/TowerBlock.pd
Windspeed Dataset
No full textThe Windspeed dataset is constructed using data from ECMWF’s ERA5 over a region of west Kazakhstan (lat 46-54, long 46-54). Hourly values for wind (u and v components), temperature (at 2m) and surface pressure were collected between 1st January 2020 to 31st Dec 2024.
The task is to predict the average wind speed in the central 50% of the region from an input of 6 hours of temperature and pressure data covering the whole region. Each input data point has shape [12, 32, 32] (6 hours x 2 variables for the channels, and 32x32 pixels forming the spatial region). The label is the mapped onto a fine-grained version of the Beaufort wind scale (120 instead of 12 classes). To avoid classes with no samples, we filter out data to only include samples where the label sits between class index 20 and 40 (corresponding to the range 2.36 - 6.69 m/s).
Data from 2020-2022 form the training set, while 2023 is the validation set and 2024 the test set
Interviews with Comedians
No full textThis dataset comprises three interview transcripts with comedians and voluntary responses from 11 comedians to a semi-structured, online questionnaire, conducted and collected as part of the practice-led research for the PhD project Disrupting Live Music: A Practice-Research Portfolio of Musical Pranks and Performance (University of Edinburgh, 2025). The study investigates the intersection of live music composition and alternative comedy.
The interviews were conducted between the 7th and 17th of November 2023, with comedians Max Boyce, Phil Jarvis and Sean Morley. Each interview probes their creative processes, audience interaction, improvisation and scripted material, failure, and the potential translation of comedic techniques into musical practice. The dataset provides qualitative insights into:
- Comedic writing and performance strategies
- Audience dynamics
- Cross-disciplinary influences between comedy and music
- Personal and cultural contexts shaping artistic identity.
All transcriptions by the researcher (Ethan Davies).
The questionnaire was sent to a number of comedians, with voluntary answers by 11 comedians collected between 2023-10-26 to 2024-06-08. It explores:
- Creative processes
- Audience dynamics
- Failure and adaptability
- Cross-disciplinary insights
- Rhythm and timing
Respondents represent diverse comedic styles, including alternative, character, physical, musical, and prop-based comedy. Ethical consent was obtained, and responses remain unedited.
These materials informed the composition of the portfolio’s final project, Ar Wawd/Avant-Laughs (2024), which merges stand-up comedy with live composed music.
Licence: CC-BY 4.0.Dataset is a .txt file that includes all appropriate interview transcripts, questionnaire questions and answers.
Contact information and other private or confidential information has been omitted.
Materials and persons referred to in the interviews are listed in brief at the end of the transcripts
Stroke and Small Vessel Disease Imaging Evaluation Forms
No full textThis collection comprises a set of structured, standardised visual rating forms developed to support the visual evaluation and quantification of acute stroke lesions and cerebral small vessel disease (SVD) burden on brain MRI or CT, primarily for research use but also adaptable to clinical settings. Earlier versions informed the STRIVE-1 (Standards for Reporting Vascular Changes on Neuroimaging) criteria [1] and these current versions capture the most recent STRIVE-2 criteria [2]. Designed in, and refined through, many cross-sectional [3-5] and longitudinal [6-9] studies including clinical trials [10,11], the visual rating forms facilitate consistent data capture across multiple timepoints- diagnostic, baseline, and follow-up imaging. They include detailed fields to record scan information (e.g., date, MRI sequences completed), and systematically assess the index stroke lesion, including hemisphere, visibility, shape, cavitation (if applicable), lesion dimensions, and other characteristics. The forms also enable comprehensive assessment of chronic and incidental pathology, including old lesions or lacunes, white matter hyperintensities (graded using the Fazekas scale), enlarged perivascular spaces, cerebral microbleeds, basal ganglia iron deposition, superficial siderosis, and global or regional atrophy. Two additional appendix forms are included to document cases with multiple index lesions or to track new lesions appearing during longitudinal follow-up. These forms aim to promote reproducibility and harmonisation in neuroimaging-based stroke and SVD research, supporting structured and consistent data collection and analysis across sites and studies. They also provide a reference standard based on visual scoring from which numerous computational lesion analysis methods have been developed and against which they have been validated.
References
1. Wardlaw JM, Smith EE, Biessels GJ, et al. Neuroimaging standards for research into small vessel disease and its contribution to ageing and neurodegeneration. Lancet Neurol 2013; 12(8): 822-38.
2. Duering M, Biessels GJ, Brodtmann A, et al. Neuroimaging standards for research into small vessel disease-advances since 2013. Lancet Neurol 2023; 22(7): 602-18.
3. Wardlaw JM, Doubal F, Armitage P, et al. Lacunar stroke is associated with diffuse blood-brain barrier dysfunction. Ann Neurol 2009; 65(2): 194-202.
4. Stringer MS, Blair GW, Kopczak A, et al. Cerebrovascular function in sporadic and genetic cerebral small vessel disease. Annals of Neurology 2025; 87(3): 483-98.
5. Staals J, Makin SDJ, Doubal F, Dennis M, Wardlaw JM. Stroke subtype, vascular risk factors and total MRI brain small vessel disease burden. Neurology 2014; 83: 1228-34.
6. Wardlaw JM, Chappell FM, Valdes Hernandez MDC, et al. White matter hyperintensity reduction and outcomes after minor stroke. Neurology 2017; 89(10): 1003-10.
7. Clancy U, Jaime Garcia D, Stringer M, et al. Rationale and design of a longitudinal study of cerebral small vessel diseases, clinical, and imaging outcomes in patients presenting with mild ischaemic stroke: Mild Stroke Study 3 (MSS-3). European Stroke Journal 2021; 6: 81-8.
8. Clancy U, Arteaga C, Jaime Garcia D, et al. Incident infarcts in patients with stroke and cerebral small vessel disease: frequency and relation to clinical outcomes. Neurology 2024; 103: e209750.
9. Hamilton OKL, Cox SR, Okely JA, et al. Cerebral Small Vessel Disease Burden and Longitudinal Cognitive Decline from age 73 to 82: the Lothian Birth Cohort 1936. Transl Psychiatry 2021; 6(11): 376.
10. Kopczak A, Stringer MS, van den Brink H, et al. Effect of blood pressure-lowering agents on microvascular function in people with small vessel diseases (TREAT-SVDs): a multicentre, open-label, randomised, crossover trial. The Lancet Neurology 2023; 22(11): 991-1004.
11. Wardlaw JM, Woodhouse LJ, Mhlanga I, et al. A Randomised Clinical Trial of Isosorbide Mononitrate and Cilostazol for Symptomatic Cerebral Small Vessel Disease: The LACunar Intervention Trial-2 (LACI-2). JAMA Neurology 2023; 80(7): 682-92
Videos of eMouseAtlas Models: Theiler Stage 16 (9.5-10.75 dpc)
No full textA number of videos for each of the eMouseAtlas 3D mouse embryo models to show the overall form and in some cases selected anatomy. Each video is identified by the unique EMA ID with annotation if required. The videos labelled as "watermovies" are captured using the OPT system with the embryo spun on a longitudinal axis with no tissue clearing