- Recognition of handwritten characters by neural Networks has been researched for many years. The most complex neural network was developed by Fukushema in 1980, a hierachical, multilayered ANN, this should be included in the paper.

- it woulde be nice to add a section , discussion of the errors made by the ANN

-the authors perform an experiment , but it is not clear which digits  are selected, how long was the training time etc.

-the architecture of the used ANN shoild be specified, probably the authors use 24 neurons as input layer and use the greyvalue of the pixels and 10 neurons as outputlayer

The article discusses current issues in the field of Machine Learning. Modern libraries have been used to process information.
It would be good to know where the approach described applies.