Neuranext Delivers Computer Vision AI Course at Beijing Normal University in China

Our first workshop lecture today here in Beijing focused on the conceptual and mathematical bases of Deep Learning. I feel it’s important to start with the fundamentals and build from there. The afternoon lab session, where students write a simple feed forward neural network in Python, really ties everything together; and the benefits of concepts like regularisation and ReLU neurons become very concrete.

The second day of the Deep Learning workshop here in Beijing, with me teaching about Computer Vision, Convolutional Neural Networks and the all important bias-variance tradeoff. It was satisfying to see the students smiling and nodding a silent approval when, in the afternoon lab session, they saw that the theoretical principles taught to them in the morning were borne out in their modelling. I’ve been very impressed with the quality of the graduate students in education and psychology here at BNU, most of whom have been well trained in social science statistics.

Finished up the Deep Learning workshop by having the students train a word embedding using Tomas Mikolov’s Word2Vec (CBoW) model. They then transferred this to a simple neural net for Automated Essay Scoring. The importance of “transfer learning” was further underscored by them training SqueezeNet on CIFAR-10 both from scratch and with ImageNet weights. The day ended with a photograph in front of the BNU’s bell sculpture, which is apparently a gift from alumni thanking the university for producing so many outstanding teachers for their nation.