i went through all the videos of the MP neuron and Perceptron including the hands on videos. I have corrected my code wrt to the errata published. However, an important step in the data preparation i.e standardization of data has been missed. Is it a miss or did the prof say that it is not required?
i went ahead and did the standardization and got good results in my first run (epoch = 1) but eventually after 1000 epochs and learning rate adjustments, the results were almost similar to what prof got in the videos. Except the weight matrix did not jump from -15000 to 25000 but for me it varied from 40 to -20.
Hence, is standardization important?