Machine Learning

We offer a range of support for researchers looking to apply machine learning techniques, from the conception of new lines of research to running large-scale models on a high-performance computing cluster on campus, or in the cloud.

Conception

We believe it’s important to have an honest discussion about expectations and limitations before we embark on a new data science adventure. We’re excited about machine learning, and believe that current approaches have much to offer for social science research. At the same time, we’re also aware of the limitations set by the machine learning models, available data, etc.

While the advancement of artificial intelligence in popular culture often seems to be almost indistinguishable from magic - the reality is that we are a long way from a ‘general’ artificial intelligence. That means that (for now, at least) whether your research questions can be answered through the application of machine learning models depends heavily on the type of question, the quality and amount of training data, and the available resources.

We’re more than happy to discuss what machine learning applications may be possible for your research, and what kind of questions - and answers! - you can expect.

Preparing a model

The success of a machine learning model depends heavily on the care and attention with which it is implemented. Data often needs to be cleaned and reshaped, variables may need to be reshaped or combined to create useful and meaningful features, and model parameters will need to be tuned for best results.

Each of these steps relies on both technical expertise as well a deep understanding of the subject matter. As such, we believe it is important to have an ongoing conversation between BDSi data science experts and researchers throughout a data science research project.

Depending on your needs, BDSi can act as a consultant for particularly complex models, or help you think through your entire model from design to implementation.

Running a model

BDSi has access to the high-performance computing cluster hosted by the EEMCS (EWI) faculty, allowing us to run large machine learning models on-premise. Alternatively, we can also help you run models in a virtual research environment such as the ones offered by LISA.

Interpreting model & results

In many cases, being able to interpret the model is as important as the accuracy of the model itself. We can help you get an intuitive understanding of the model and its’ results, and create visualizations to communicate with your audience.