Building and Optimizing a Gradient Descent Machine Learning System
In this meeting, Camir is seeking guidance on building his own gradient descent machine learning system. He is unsure how to start and is looking for help on the function stack. Ray provides guidance on building the backend technology similar to Microsoft Excel and using a while loop to test various options. They go through the Excel document and analyze the input variables, object tranches, and the desired output, d25. They create a function that iterates through tranches and adds shares to the in the money shares if the exercise price is less than the share price. They also calculate the in the money dollars by multiplying shares with the exercise price. They then calculate the repurchase shares by dividing the in the money dollars by the share price. Finally, they calculate the fully diluted shares by adding the repurchase shares to the in the money shares and calculate the implied share price by dividing the equity value by the fully diluted shares. They create an error variable by taking the difference between the share price and the implied share price. Afterward, they create a solver function that implements a while loop to iterate through the game of finding the correct share price using the gradient descent method. They track the number of loops and set a condition to break the loop if the error is below a certain threshold or if the count exceeds 1000. They return the share price as the output of the solver function. Camir expresses his gratitude for the guidance and states that he feels enabled to build on Xano. The meeting concludes with the participants scheduling another meeting for the following day.