PLEASE UPDATE with TRACK CHANGES
VERSION 1 – Wouter Haerick
UPDATE 1 – Tom Dhaene
UPDATE 2 – Dirk Deschrijver
UPDATE 3 - Ivo Couckuyt…
UPDATE 34 -– …Joni Dambre
Input for IDLab website:
Machine Learning & data mining
1. Topic Title
Machine Learning & Data Mining
2. Short text
Target Length: max. 2 lines
State-of-the- art machine learning tactics techniques for image, video and sensor data processing (neural networkss, gGaussian process regression and latent variable modelses, re-servoir computing, surrogate metamodels, …)
3. Long text
Target Length: max. 20 lines
Intro-text: 2 lines - why this domain is important nowadays
The volume of data, that companies are gathering, is increasing rapidly in. Traditional data analytic tools hit their limits. They suffer in processing realtimereal-time data streams, and do not easily scale towards big data sets. Novel algorithms, tools and platforms, such as SPARC, HADOOP, KAFKA, HIVE… however open new opportunities.
IDLab Subtopics: 10 lines – text with subfields we address and why we are well positioned
At IDLab, we offer the full spectrum of machine learning and artificial intelligence expertise. This covers both supervised learning (classification, regression, …) and , unsupervised learning (clustering, data pattern recognition, …) and reinforcement learning. . DNeural networks, reservoir computing, dDeep learning neural networks for image, video, audio and sensor signal analysis/classification, probabilistic generative models, and deep reservoir computing, Q-learning for robotics and game applications are only some of the techniques we are mastering.
Our experts tackle the most complex data challenges, and differentiate with automated machine learning solutions. These solutions apply machine learning techniques to optimize the hyperparameters and to select the best machine learning model. and to optimize the model parameters.
Datasets are often not so well suited to apply machine learning straight away. At IDLab, we have built the expertise to facilitate dimensionality reduction, outlier detection, mitigation of missing data, denoising of data, facilitate feature extraction, dimensionality reduction, outlier detection, mitigation of missing datamodel selection and hyper parameterization, and sensitivity analysis, denoising of data.
Examples: 8 lines one or two examples
The data scientists at IDLab are internationally recognized for their machine learning insights and capabilities.
* In 2016, a team of IDLab scientists finished among the winners of US 2nd National Data Science Bowl (on Kaggle.com), with a machine learning algorithm to automatically determine cardiac volumes from MRI scans. In 2015, the this team also won the “Deep Sea” challenge , out of 1049 teams, in the 1st National Data Science Bowl.
* The surrogate modeling (SUMO) toolbox for data-efficient machine learning, developed at IDLab, is worldwide recognized worldwide as state-of-the-art, and has been downloaded more than 3000 times across the globe. Licenses have been requested by players in automotive, bay-area based semi-conductor companies, etc. The toolbox can be used as an add-on to the popular simulation and design automation platforms (ANSYS, …).
At IDLab, the data scientists from different labs, previously known as MultimediaLab, reservoir Reservoir computing lLab, data Data science Science labLab, and surrogate Surrogate modeling Modeling labLab, are joining forces.
4. Picture:
Add image that illustrates the field / demonstrates technologie.
Adapt the short texts: