Data Science Course

We provide different training opportunities based on your data science need and availability

We run short on-demand ad-hoc lectures on machine learning and data science targeted toward specific School, Institute or research group. The format is 1 hour of introduction to data science or a specific topic of interest followed by 1 hour for Q&A. If you would like us to organise one, please get in touch.

We also run a day-long data science course (Day of Data Science) with multiple talks, hands-on sessions, and Q&A designed to give you a broad overview of state-of-the-art data science tools and how they can be incorporated in your research.

Upcoming

We are developing an introductory course on data science available without cost to the University staff. The planned format is five 2 hour sessions with lectures and hands-on exercises. Please get in touch to learn more. Find a tentative course structure below.

Course Structure

Data handling
1. How do you deal with missing data, censored data, outliers etc?
2. How to deal with features? e.g., raw data versus feature engineering.
Exploratory analysis
1. What model might be useful? Ordinal regresson versus classification, linear regression versus mixture of linear regressions
2. What question can I ask? Hypothesis driven versus exploration driven
3. How are my features distributed? Gaussian versus non-Gaussian, categorical versus continuous, positive or zero inflated
Modeling
1. How do you handle class imbalance? up/down sampling, weighted loss etc.
2. How do you exploit unlabeled data? semi-supervised learning versus supervised learning
3. How do you exploit multiple views? multitasking versus multiple unitasking
4. How do you use meta-information?, e.g., second level information in multi-level modeling.
5. How do you select the best model? meta-learning versus ensemble learning
Evaluation and diagnostic
1. Is your model correct? Model diagnostics and criticism.
2. Is your solution stable? evaluating sensitivity and quantifying uncertainty in the solution.
3. Is your method working? Evaluation metrics. Accuracy versus AUC. Precision-Recall versus ROC.
Visualization and interpretation
1. Is your conclusion correct? p-values and false detection rate correction.
2. How do you interpret the weights? raw values versus variance normalized value.

This article was published on 2 May, 2022