Hello 
I’m new to neural networks and working on a classifier that specifies the health of humans based on different properties. The properties and the actual state of health are given data. Goal is to give a accurate forecast with given properties (without the actual state of health of course). The basic neural network is done with an accuracy of 88%. How could I optimize this value?
Thanks for your help 
Hi theropoy!
Your question is very broad. (As it stands, it’s not specific to
pytorch, so it might be off topic, but I’ll let others decide that.)
Folks can probably give you some guidance if you give us more
detail about what specific problems you are facing.
Generally, what does your data look like? How many “properties”
are there, and what are they? Are most of them specified for most
samples? What is the “state of health”? Is it a single score
(“good”, “bad”, “six feet under”) or does it have several dimensions
(heart disease, cancer, etc.)?
How many (labelled) samples do you have?
Could you show us the (pytorch) code for the network you have
trained?
There may well be things you could try to improve the accuracy
of your network, but without details about your problem, there’s
really nothing we can suggest.
Best regards.
K. Frank