Null Inputs/Inhibiting Inputs & Outputs with Scikit-Learn MLPRegressor

I'm trying to build a general predictive model of a model of a machine.

I've got a variable number of sensor inputs, and I'd like to create a MLPRegressor that can estimate outputs from the input values.

I know I can create individual AIs to model each individual output (ie. if I have 5 inputs, I can make 5 different AIs with 4 inputs each). But given that I have some large numbers of inputs, I was hoping for a more generic, single-AI solution.

It would be great to have a single network with the full number of inputs and outputs. But of course, it would result in the network simply copying the input value to the output value.

Ideally, I'm looking for a way to create an AI that won't connect matching input output indicies.

One solution I would be interested in trying is inhibiting specific inputs and outputs during MLPRegressor training and fitting.

For example, you could have a 5-input, 5-output network and train it 5 times. The first time would use inputs A,B,C D, and only output E. The next would be A,B,C E and only output D etc (in practise these would need to be shuffled together so you don't over-train the network).

Then, if you wanted to solve for 'C', you could inhibit input 'C' and fill in the data for all of the other inputs. This would give you an output for 'C' that was based on all of the inputs that aren't C.

Or if there's some other way to achieve the same sort of outcome of making an AI that won't connect matching input/output indexes?