Silas is a generic data mining and predictive analytics software built upon advanced machine learning, automated reasoning, and artificial intelligence techniques. It can deal with any type of structured data and it can be used to perform tasks such as classification, regression, segmentation, anomaly detection, prediction, etc.
To achieve fast computation speed and low memory usage, our high performance computing engineers have chosen to write the core code in C++. Silas is built to ultilise all the computation power possible using advanced multi-threading techniques. We have used program analysis tools to verify our source code and ensure that every resource is released from the memory as soon as possible and there is no memory leakage.
Besides high performance, Silas has two unique features: The first feature is a machine learning model analysing tool called Model Insight. This tool adopts the latest automated reasoning technique to provide insights on the model such as the rationale behind the model’s decision-making, key features used in the prediction, controversies between sub-models etc. The second feature is called Model Audit. This feature uses formal verification techniques to mathematically prove that the prediction model conforms with the user’s specifications. The user can also re-train models with pre-defined constraints and ensure that the models are safe, correct, and performant.
Silas Edu, which is a spin-off of Silas, is a command line application featuring the tools to train, audit and introspect binary classifiers built from tabular data. It is FREE for non-commercial use only.