Two Issues with Neural Network Structures #32
Description
Hello, I have recently read your paper and code carefully. Your ideas are very ingenious, and the code indeed reproduces the experimental results. However, I have the following questions:
1.Based on my understanding, the network you designed has only one layer of depth. With such a level of complexity, can it effectively capture the features of more complex reactions? This is particularly concerning in cases like Case 3, which involves enzyme reactions. The nonlinearity and complexity of such reactions should be quite high.
2.I understand the elegance of representing the stoichiometric matrix through network weights. I also appreciate your incorporation of the law of mass action and Arrhenius law. However, from the perspective of the network structure, it seems indistinguishable from a traditional neural network. To some extent, I believe this network hasn't incorporated any special treatments; essentially, it still follows the form Y=σ(wx+b), with specific activation functions chosen. I’m not sure if my understanding is correct.
Thank you for your response.