Skip to content

Latest commit

 

History

History
32 lines (28 loc) · 2.91 KB

File metadata and controls

32 lines (28 loc) · 2.91 KB

ucl-machine-learning-projects

Courseworks including jupyter notebooks and PDF reports from courses of my master's at UCL. The files included are:

  • deep_representation_learning/

    • concept-bottleneck-age-brain-mri.ipynb : adopting a concept-bottleneck modelling technique for interpretable prediction of age from brain MRI scans
      • some interesting things I tried was using Bayesian optimization for hyperparameter search.
    • cb-age-prediction-brain-report.pdf: associated report to the above notebook
    • review-antibiotic-discovery review_antibiotic.pdf : a review of the paper on A deep learning approach for antibiotic discovery.
      • you can read it as a short tutorial on how graph neural networks work
    • temperature-prediction-RNN.ipynb : using a simple RNN for predicting temperature from the classical Dehli climate dataset
    • function for auxiliary loss used
    • convolutional-VAE.ipynb : using convolutional VAE architecture from this source with the loss function from this source for MNIST digit reconstruction
      • the only thing worth noting is to not use the Bernoulli distribution before binarizing MNIST digits.
  • hospital_readmission/

  • biomedical_imaging/

    • diffussion_mri and Figures-dmri.pdf
      • report with associated figures on fitting linear models to brain diffussion MRI data yielding insight into the brain's structure
      • for more information on methods, please watch https://www.youtube.com/watch?v=VyZfhkyKFNQ
    • biomed_imaging_modelling_connectome.pdf and Presentation -CW3 Brain Connectivity.pdf
      • report and associated slides exploring simple linear functions modelling the brain's structural with functional connectome (which are essentially pairwise correlated voxel values between cortical regions of the brain in each respective MRI modality)
  • msc_mcmc/

    • mcmc_multispecies_coalescent.pdf: written report on using MCMC for estimating common ancestor population size and speciation time for the human-chimpanzee modelled with the multispecies coalescent model
    • mcmc-msc-vectorized-submit.ipynb: code implementation of a vectorized Markov Chain Monte Carlo used for parameter estimation for the multispecies coalescent model for two species (human and chimpanzee)