During our scientific education, we were trained to discover and understand what nature seems to hide from us: observe or measure, create a model with your understanding of the principle behind it, fit the data to the model, evaluate the model, improve it and iterate. With the large amounts of data that we generate we need a computer and a computer based version of our model to do so: an algorithm. Algorithm development has been one of our core competencies since the foundation of cosine.
Artificial Intelligence might have a kind of mysterious ring to it, but at its core there are principles that are congruent with what physicists have been doing for decades: you create a model (algorithm) of the interpretation of reality, you test it with data, and you use the errors in the data or in the results to correct your algorithm parameters. Having that said, sometimes we are still surprised how fast a machine learning approach sometimes yields the desired results, especially in what we call a “hybrid approach”, where we feed the model with the physics we know applies and train the model with data that is labelled.
Contact us to inquire what we can do for you!
cosine has extensive experience in developing, at times in close collaboration with our clients, data processing software. This ranges from low-level embedded software to top-level visualization and control software. cosine has done this at a scale ranging from single person projects to managing large teams of dozens of developers.
Advanced measurement systems can generate large amounts of data. An important aspect is the storage of this data in a structured, standardized way that allows for easy retrieval now and in the future. Given the large amounts of data and different levels of processing, it is important to distinguish different types of data and adjust the storage strategy accordingly.
Image analysis is the extraction of information from digital images through the application of digital image processing techniques. Computers are indispensable for the fast analysis of large amounts of data with complex algorithms. Still, when thinking about an image analysis problem for the first time, we are immediately confronted with the dichotomy between the ease with which the human brain can identify and recognize shapes, and the difficulty of devising an algorithm able to reliably perform even the most basic geometrical shapes.
