MNIST Digits Sum Network | Dataford Interview Questions

Build a deep learning network that sums single digits from the MNIST dataset; the summation must be performed by the network itself.

Implement build_digit_sum_network(input_shape) using only the Python standard library. Return a deterministic architecture specification for two grayscale MNIST images per sample. The specification must use one shared digit encoder, combine the two encoded representations, and produce one scalar output per sample without an explicit arithmetic addition operation.

input_shape is [batch_size, 2, 1, 28, 28]. Return a JSON-compatible dictionary containing the input shape, output shape, shared-encoder flag, explicit-addition flag, maximum supported target, and ordered layer names.

Constraints

input_shape has exactly five integers
input_shape must have the form [batch_size, 2, 1, 28, 28]
1 <= batch_size
The network supports two digits per sample
The largest possible target is 18
The returned value must contain only JSON-compatible data

Function Signature

def build_digit_sum_network(input_shape):

Your question is MNIST Digits Sum Network. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.

Problem

Build a deep learning network that sums single digits from the MNIST dataset; the summation must be performed by the network itself.

Constraints

input_shape has exactly five integers
input_shape must have the form [batch_size, 2, 1, 28, 28]
1 <= batch_size
The network supports two digits per sample
The largest possible target is 18
The returned value must contain only JSON-compatible data

Function Signature

def build_digit_sum_network(input_shape):

Your question is MNIST Digits Sum Network. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.