Haiku Tech Center

I was planning on continuing our exploration of transfer learning and our old friend VGG16, but this was just too interesting not to pass up.  Keep in mind that i'm probably more excited about the general concept we're going to discuss, as opposed to being blown over by the actual system presented. So, there's an interesting article out today on the 3dPdModular modular synthesizer.  It's a euro-rack like modular synth, complete with different modules that plug into a chassis and patch cords to hook things up.  But inside it's all digital. And it's all based on running Pure Data (Pd) on a Raspberry Pi. At this point some of you might be shouting, but the entire point of a euro rack modular synthesizer is to utilize all of those wonderful analog modules and their mythical unique analog sounds.  Yeah sure. But as we have mentioned here before, the climate on Maui is not really amenable to fine precision analog hardware, as it instantly begins to rust and

Typically when you think of applications of deep learning neural networks, they include the types of things we've been discussing here, like image recognition, audio analysis and synthesis, language completion or translation, etc.  problems usually thought of from the standpoint of statistical modeling or manifold learning. You don't typically think of them for use when dealing with symbolic math problems. I recently came up on an interesting paper that addresses this problem. Guillaume Lample and Francois Charton in Facebook's AI research group proposed an interesting way to restructure symbolic math expressions so that they could then be used to train RNN seq2seq neural nets .  They restructured the symbolic expression problem so that they could use the same deep neural nets used for language translation. What they did was restructure mathematical expressions as trees. Operators and functions are internal nodes, operands are children, and numbers, constant and varia

Todays HTC Seminar Series talk rounds out our presentations (#3 of the godfathers of deep learning) by presenting a talk by Yann LeCunn of New York University and Facebook. The talk is the Cosyne 2019 Keynote Lecture on 'Feedforward networks, deep and shallow'. This is a fun talk that covers a lot of ground. It's a good introduction to the history of deep learning research.  With a lot of good tie ins to human visual system computational modeling work. It also spends a lot of time discussing where deep learning research could go next.  So a discussion of unsupervised learning, energy approaches to representation learning, sparse coding, etc. Lot's of interesting ideas discussed. Mask R-CNN is also discussed.  Its available as a PyTorch model, and we'll take a look at it in a later post.

VGG16 is a deep learning neural net.  Developed for image classification. It's a convolutional neural network model proposed by K. Simonyan and A. Zisserman from Oxford University.  You can read all about it in their paper  “Very Deep Convolutional Networks for Large-Scale Image Recognition”.  Here's a good overview of VGG16 on neurohive . People care about it because the model achieved 92.7 % top five test accuracy for the ImageNet test. ImageNet is a training set of labeled images put together to use as a benchmark for  different computational approaches to image classification.  It is composed of over 14 million labeled images belonging to 1000 classes.  The images were collected from the web and then labeled by human labelers working for the Amazon Mechanical Turk crowd sourcing tool. VGG16 assumes the input image is of size 256x256.  It also has some assumptions about data normalization in the 256x256 input image.      All of the models we discuss use 224x224 for t