Some of us at the Intuendi team just came back from the first Optimization, Big Data and Applications summer school.
Among the course teachers, Chih-Jen Lin, author of libsvm, and Saathiya Keerthi (Microsoft research), who opened for us the secrets of Generative Adversarial Networks perhaps the hottest research topic today in data science.
The connection between Optimization and Machine learning is that some of the most important machine learning methods as Neural Networks and Support Vector Machines (SVM) have training processes based on optimization algorithms.
However due to the high volume of stored data, traditional optimization methods are not the best practice. In fact, they take a long time before finishing. Hence, in the last few years, the scientific research has been paying attention to Big Data techniques.
During the summer school, efficient optimization methods were proposed in order to deal with large datasets. In fact, all the teachers proposed new methods for handling datasets with high number of features or/and examples. Both the teachers Chih-Jen Lin and Saathiya Keerthi are also main contributors of LibLinear, an effective tool which uses efficient optimization methods for training Linear Support Vector Machines.
Machine Learning methods are also used for demand forecasting. Their goal is to extract some not trivial information from the history in order to improve the forecasting accuracy and to save unnecessary costs.
Although it is too early to put this stuff in a product, we at Intuendi look at these advances with great interest, as they might boost the quality of demand forecasting by exploiting hidden variables in our time series predictions. We at Intuendi already use machine learning in the forecast engine and we are always open to current research, looking for any possibility to make our engine better and better.
