Areas Nitin Williams is Knowledgeable in:
analysis of Neuroscience data, Machine learning
Techniques Nitin Williams Uses:
Pattern classification: Logistic regression, Support Vector Machines (SVM), Artificial Neural Networks (ANNs)
Statistics: Multiple regression, Linear Mixed Model, ANOVA, ANCOVA, t-tests
Multivariate analyses: Structural Equation Modelling, Hidden Markov Modelling (HMM),
Multi-Variate Auto-Regressive (MVAR) modelling, Independent Component
Analysis, Principal Component Analysis
Grouping: k-means clustering, Empirical Mode Decomposition (EMD)
Complex systems: Graph theory, measures of causal interaction (e.g. Transfer Entropy), measures of functional interaction (e.g. Correlation, Coherence, Mutual Information)
Optimisation: Genetic algorithms, Simulated annealing
Nitin Williams's Problem Solving Skills:
- Data analysis
Nitin Williams's Problem Solving Experience:
- Developed pipeline based on MVAR modelling to delineate functional
networks from task-free MEG data. Found differences in network properties of
younger and older participants.
Williams N, Henson R, Taylor J (2014) ”Measuring effective connectivity in
resting-state MEG using PDC: effect of ageing in the Cam-CAN project” OHBM
2014, Hamburg, Germany - Developed method based on HMM to characterise EEG task-related networks
as they change across time. Demonstrated ability to discriminate between trials
from two conditions based solely on HMM parameters. Presented work as
invited lecture at OHBA, University of Oxford, UK. - Obtained 70 % single-subject classification accuracy using logistic regression,
discriminating brain activity patterns while participants viewed images of faces
or houses. Electro-corticographic (ECoG) data was used. - Devised novel measures to diagnose asthmatic patients based on capnographic
data. Simple t-tests on certain features revealed statistically significant
discrimination between asthmatics and controls. - Used EMD to perform filtering of task-related Electroencephalographic (EEG)
Decomposition activity. Compared this technique to other forms of filtering (e.g. low-pass filtering) and found superior performance of the EMD-based method.
Williams N, Nasuto S, Saddy J (2011) ”Evaluation of Empirical Mode
Decomposition for event-related potential analysis” Journal of Advances in
Signal Processing 2011:965237, pp. 1-11 - Developed a pipeline based on k-means clustering, for grouping ensembles of
task-related EEG trials into sub-types. Validated on simulated & expt. data.
Williams N, Nasuto S, Saddy J (2015) ”Method for exploratory cluster analysis
and visualisation of single-trial ERP ensembles” Journal of Neuroscience Methods
250(2015), pp. 22-33