Python-based Deep Learning Solution

I'm seeking a qualified freelancer to work on a Deep Learning and Python-based project.

Description:
Use the classic MNIST handwritten digit dataset in the attached code file.
Report performances of three (3) different models by changing
a) activation function (relu, sigmoid, softplus, tanh, elu, selu, etc.)
b) batch size and number of epochs
c) optimizer (e.g., SGD, Adam, RMSProp, etc.) and learning rate
d) Number of layers (e.g., 1,2,3,..)
e) network architectures: number of neurons in each layer (e.g., [64, 32, 16], [16, 16, 16])
f) using early stopping (change patience number, change validation set size)

Collect the performance and compare them
Any recommendations for the hyperparameter tuning?

Submit a notebook containing three (3) working models and their results. Accuracy of those models is not critically important for this project.

Skills and Experience Needed:
- Proficiency in Python
- Extensive knowledge in Deep Learning
- Demonstrated expertise in creating and training deep learning models
- Solid experience in code implementation
- Ability to write comprehensive reports and documentation

Potential freelancers should provide as much relevant information as possible including past work, project-related experiences, and a detailed proposal showcasing your understanding and approach to the project.