Breakthrough Machine Learning – Generalizing loss. Transform the loss 0..1 with a model2 the loss of model1. So if the loss oscilates wildy the generalizing loss dont. Its confined between around 0..1. Can the iterate with the same input for 100000 times without degrading. By Per Lindholm

https://peroglyfer.se/wp-content/uploads/2019/12/Screenshot-from-2019-12-09-09-39-07.png

https://peroglyfer.se/wp-content/uploads/2019/12/Screenshot-from-2019-12-09-09-38-23.png

https://peroglyfer.se/wp-content/uploads/2019/12/Screenshot-from-2019-12-09-09-40-22-1.png

You can also learn if the function is going to converge or not. By random choice selecting between at=loss for converge = 1 and at= np.random()*loss for converge = 0 Then a model2 learns the converge label. Voila I guess the Halting problem is solved?

Machine Learning Idea – There are alot of zeros in machine learning target. So what if you can make patterns in the zeros. Assume some kind of zeroToSmallValue Pattern is beneficial to the classification. // Per

Machine Learning – Generalizing Idea – Sometimes convolutions produce shadows. Shadows from different objects can appear the same. I think you also get generalization from ||derivative similarity||. With a some small number of value similarities. // Per

Safe Affordable Life – Innovation – Wash Your Clothes Securely at the workplace. #Africa#Sweden#US

Electric Airplane Energy Strategy – Since electric motors and battery are good at acceleration. Could a strategy be to accelerate faster the plane for longer periods of time. Accelerate – glide – accelerate? for electric Cargo Planes?

?Lindholm Convolution – Inspired by random shape images. I wondered what x belong to an image which pixels colors are assumed to be random. cumsum(x-mean(x)) = image. I think I found a convolution style. // Per

Association Intelligence – Random numbers has something to do with shapes and figures. Since you sometimes (randomly) can recognize shapes from a set of random numbers. ||Organic shapes are so difficult that they are adapted towards random numbers||. superWow! // Per

Math Idea – Looking at Schrödingers equation. Every topic in math should be able to exist in a differential equation. Geometric diff. eq. Statistic diff. eq. Algebraic diff. eq. Programatic diff. eq. ?Analytic diff. eq.

Remember Einstein: He made classification problems. Cold fusion could be a property light emission problem. Put a different atom layer between two equal atom layers. That’s enough for a classification problem//P

Machine Learning Physics – Guess – Continuity Fusion? – I assume there exist a property effect when you got three layers with another atom type in between. Apply energy and does not want to interfere with the light emission. Otherwise a classification problem. -> Property fusion.

Machine Learning Physics Guess – With Quantum Loss functions maybe there are also Quantum Fusion. Since there is so much energy. A smart way would be to extract energy on a small scale. Quantum Fusion could be similar to glow plugs I guess.

Purpose of Materials give Innovation Clues – Guess – I wonder. If you know what the planet and ocean uses the material for. Could this information provide clues where we could use this material. By viewing the materials or atoms contact history. (in water?) // Per

Machine Learning Physics – Materials are models() and as such they could store knowledge in the form of loss functions. An advanced universe will have figured out how to store a lot of loss functions without losing information. Oil, water, wood, rocks, air.. all carry information

Simulating 77 with a list of target data points and a model in between. Difference between absolute error and squared. Both are ?not suitable for diff eq. Maybe a diff eq is a perticular loss function diffloss.

Machine Learning Engineering – There is another way to think of the world other than in differential equations. Using a x = model([1,0]) as the position. You only need a list of targets a list of xt values. Then when the first target is met you change to target two.

Machine Learning Physics Guess – Could 1/(x+273.15) for x in Celsius degrees exist in some physics equation since you cant go lower than the absolute limit. There must then exist a singular equation with the temperature. // Per ( Like in black hole )

Car Innovation Idea – Instead of making small animations or still images of the style of car. Recognize which movie they would fit into and make something of that. That is Machine Learning deep fake replace cars in movies to see which style fits in.

Car Innovation Idea – Inspired by Tesla CyberTruck should brands have their own Fantasy, Science Fiction Style Cars? I mean its electrical which leaves imagination to fill in the spaces.

Super Desalination Guess ? Like in machine learning you have a target. Could you therefor change the light refraction/reflection index result just one bit from salt water ?1.337 to freshwater 1.330. Would this mean that the saltwater knows what to do?

Thats how to do it I think. Machine Learning target of light emission or light indexes. You name it. All that are outputs of mass are targets. Which you indirectly affect so it can self adjust. Wow

Refugee Shelters replaced with Tiny Houses? – Create Assets that can be sold – Work and pay rent? – Rent to own. Create lots of money. Reuse solutions that exists already. Split Air conditioners? Heat burners. Toilet. Water. Its the Biggest Win Win // Per

Tiny house on wheels since you can have them moved to new locations

Quantum Information Transfer – I wonder if quantum information transfer is |data to a loss function|. Since you normally don’t update the input with the output the transfer is in one direction. If so we can measure loss functions to solve many problems. // Per

Short Cut Way To Fusion Energy – Change the refraction index towards helium of a point target. The idea is that the machine learning loss function depends on the output light emission. So ordinary hydrogen to helium fusion could be possible with very small amount of energy. //Per Wow

Peace Innovation – Countries involved in peace keeping in Syria must have a peace minister in their government. Otherwise that government will be overly biased towards war and not peace. Its like a self regulating feature of our civilization. We must be smarter when are not. // P

Mega Idea – Other types of energy? Carbon-Electric energy probably is much more intense than ordinary metal-electric energy. For this you might need a grafite battery? I assume you get carbon-electricity from say hydro to carbon-turbine-generator. Classified coming from carbon.

With electrons coming from carbon. With Carbon-Electric-Energy maybe you get less internal friction or losses.Needing ?much higher voltages