Food innovation – Try the set of possibilities. Fast Noodles + boiled water and later after 6 min add milk to cool it off and give it more flavor. // Per
Machine Learning – Hot encoding is not statistically ?friendly. I think you need the target to be a double normal distribution. Around 0 and around 1. For this you can add yt = yt +m ; m = model2(X) and then with subtract m from the model(X_test) – m.
Important Machine Learning Human Thinking – Move the body while your thinking. Move your arms, talk loud etc. Why? Since every mass unit in the body wants to improve the loss value set by the brain. There for everything adapts to find the answer. // Per
Machine Learning Guess – Could letting a layer act as incriment at the index i,j and multiply/add? with value for the next layer?
Guess – In machine learning math there is a connection between complex numbers and variation. So in F.sum(X) numbers. They to have variation in X. So there is a possibility for more variation in matrix multiplication other than F.sum() // Per
Machine Learning Battery Idea – Treat ions as if they were ML models(). Could things be added to make the ion models make better predictions. Is there battery music for better predictions within. // Per
If machine learning physics is something then antimatter is just a possibility. 0*number = 0 but what if you invent “Event numbers”. If anti number + number = anti number event.
MAchine Learning math adapts so its inventive not so much a discovery
Machine Learning Innovation – Since frequencies don’t change. They can then be used as universal layer loss function values. So each layer is not allowed to change the frequency. You use that as a limiter for each layer. Everything in the universe is enabling larger models // Per
Twitter idea – Use a machine learning model() to allow spell corrections and missing words. allowed = 0 or 1 = model(tweet). Just an indicator not if statements // Per
Machine Learning Battery Innovation Food-Mimicry – Does there exist something like food-mimicry for getting inspiration for batteries. Are there general problems that would be the same. Both are heated. Do spices have other functions in batteries. // Per
Machine Learning Electric Plane Innovation – Use machine learning model() to mitigate beam bending of the wings. To know when to use power that is. Every reltative position and event in the plane body must coincide with the motors. // Per
Machine Learning Innovation – What if .Adam or .SGD has a 97%efficiency rate. How do we model a soft update optimizer?
Math Innovation Guess – Could the continuum be augmented with free fall limited functions. Functions similar to chains. Limited freefall functions. // Per