skip to content

Department of Computer Science and Technology


  • Reduction Algorithms for Neuroimaging Data
    Contact: Kristjan Kalm, MRC Cognition and Brain Sciences Unit

    The study of brain connectivity in terms of causal interactions between neural populations is limited by the high dimensionality of data. Neuroimaging datasets typically represent a series of data points recorded from tens of thousands of units. However, information theoretic measures, such as mutual information and transfer entropy, are hard to estimate for continuous variables of high dimensions. Recently, several approximations for such measures have been proposed, such as the maximal information coefficient (MIC) and partial distance correlation (PDC). The aim of the project is to evaluate the suitability of these two dimensionality reduction algorithms on a large neuroimaging dataset (The Cambridge Centre for Ageing and Neuroscience). The results of the project should quantify how well the MIC and PDC approximations replicate the results obtained with an existing low-dimensionality (bivariate) approach.

  • Developing data-analysis algorithms to monitor and model complex insect behaviours
    Contact: Caroline Fabre, Department of Zoology

    The aim of the research is to identify and explore the underlying neuronal networks underlying insect behaviour, in particular those performed during courtship. To do so, the laboratory generates high-speed high-resolution videos of pairs of insects performing courtship. Their behaviour is monitored and annotated, and the various types of behaviour displayed by each individual can occur simultaneously and typically occur several times during each recording in a partially overlapping manner. We obtain data files containing all the events and we aim to extract as much information as possible in order to find correlations between behaviours and understand how the two individuals interact and communicate. We assay and compare different types of flies, such as mutants in particular genes of interest or flies deficient in specific sets of neurons. This will enable us to learn more about the genetics and neuronal encoding of animal behavior.

    Material already available: Several excel files containing the different behaviours annotated and events recorded for a series of insect pairs; Macros in Excel and partial R programmed code to analyse the data.

    Primary success criterion: The ability to extract information from our videos annotations and to produce figures displaying this information. This will allow us to compare different flies, such as flies with deficiencies in some neurons or some genes.