Down the rabbit hole today: Looking at Comparisons between Hebbian and Oja models for Neural nets. I poked around in the world of conversion of categoric data (words) into numeric data, it seems that Markov chains lend themselves better to text analysis than neural networks. It also strikes me that there is a lot of work still to be done around the neural weighting. It seems to me that there is a huge body of work still to be done around node responses and modelling. I also start feeling that TPM and GPU capabilities may need to be moved onto Hadoop and that JVMs should be supporting CUDA to allow for really large scale ‘nets.