Swapnil_to_Alex

• Very well thought about the usage of econometric and Machine learning in calculating TFP. Kudos for that.
• I see this project has a wider objective of estimating the correct firm-level productivity which can further be used as an application to many papers, like Hsieh and Klenow (2009), which are using firm-level productivity. I see estimating firm-level productivity using Machine learning as a more attractive project because you are even talking about using Neural networks to approximate the production function then estimate the productivity. As a suggestion, you should explore and focus on this area, and then once you have the results then applications are unlimited.
• Are you planning to estimate Psi parametrically or non parametrically? In your equation (8), I am assuming you will be estimating the unknown function without actually requiring state variable data because that would be very difficult to get at firm levels. But in the machine learning  (neural network model) you still might need that. 
• Are you planning to estimate the 2 stages jointly or separately? because jointly with the analytical method would be unreliable.
• overall great Great work. I would like to see your further research on this topic specifically the outcome of the machine learning tool.