At Semantive we value both enthusiasm to learn and eagerness to share the knowledge with others. One example of the latter is last Saturday, when Maciej Ociepka and Amadeusz Kosik from Semantive led one day workshop on Deep Learning topic.
Deep Learning is a trending topic in Machine Learning that focuses on using different kinds of neural networks to perform various regression, classification and generation tasks. The effects of the DL algorithms are often more spectacular than any other well-known ML algorithms. Various examples of successful appliance of the Deep Learning include autonomic cars, face detection and even self-taught walking robot. The whole topic of neural network is not fully discovered yet and both new ideas for networks and new use cases are popping up almost every day.
During our workshop the participants had the opportunity to get familiar with both the theory behind the whole DL (and ML) as well as the most popular frameworks for using it: TensorFlow and Keras. The challenge that we have chosen to tackle is handwritten digits classification. It is simple enough not to get buried with feature extraction and optimization topics but deep enough to demonstrate techniques on matters such as dropout or regularization.
The notebooks used for the workshop are available on our GitHub.
We hope that we were able to share our enthusiasm on Deep Learning topic with the workshop participants. From the first edition we got valuable feedback to improve the workshop and use it when planning the second part – the Convolutional Neural Networks.