Monthly Archives: June 2018

Machine Learning Speculation – Electric Circuit – If F.relu() is a Diode then what is the electric circuit equivalent of F.sigmoid() – Probability And Classification In An Electric Circuit?

Some morning idea I forgot to post from twitter.

Here is my machine learning physics electric circuit idea.

If F.relu() is a diode then what is the electric circuit equivalent of F.sigmoid()? Is there probability and classifications in an electric circuit also?

Then if moving a machine learning classification results could mimics some sort of gravity function? Since I suspect there two options. Either you move mass together or energy between them in a problem of potential.

So mass as a movable potential?

?Advanced Machine Learning Speculation – 3D Object Inspiration To Yet Discovered Problems

What if you take abstract concepts of machine learning and its components and make it more world like.

So I imagined or really stumbled upon 3D objects with a random displacement texture mapping. From this I wonder or speculate that random as a component in machine learning can be mode more complex than just a normal or even distribution.

Since they are pretty jaggidy I used a smooth modifier. I assume there are ?many problems which have smooth_random().


Machine Learning Physics Guess – Should There Exist An Energy Recognition Function?

Should There Exist An Energy Recognition Function? Like a machine learning function.

Take a video of a spinning wheel. Then calculate the energy per unit object? Like the all 3D geometry baked into the unit. I think this is possible. At least some approximations can be done.

Then could this help iterate a better battery? This from 2D or 3D model graphics?

Peace Idea – Any Type Of Probability Can Move Us Forward. Peace Will Do It 100%

Probability can move you forward, that is my starting assertion. I’m not meaning standard probability of things happening again. The frequency of events. But associative probability. The sense of probability of connected words or things are true. Like an A.I connected word function or thesaurus.

What if you reflected upon your ideas. Transform that idea process into a sequence of guesses. With that you would have gained lots of guess experience. This is a guess of its own. But it could have a high Return Of Investment. You learn to make guesses.

A probable theory can have some truth value. Since it can help generate lots of guesses with some truth connected to it.

So probability guesses can move you forward. That is. Guesses that you know are true for a single circumstance but you also use it for other things.

Easy Slum Innovation – Bus Routes To Be Able To Go To Work From Poor Areas. Increase Chances For Better Livelihood. Its Cheaper To Bring Home Food By Bus? Better For The Environment. Light-Weight Bus Wheels?

The idea is simple.

I think you can have bus routes from slum areas to cities, work and food places. The reason is that this would increase people living in slum areas better chances of livelihood.

Money can be ?saved since its cheaper to bring home food by bus. This money can then be invested in local business. Upgrading the slum.

Also it will better for the environment. I think you just need to make light-weight wheels on buses to have it running electric.

Fractal Machine Learning Idea – Geometric Seed Shape Values From Layer Weight Matrix Outputs?

Looking at face recognition where there could be shapes combining eyes, ears, mouth and facial expressions. I wonder.

Could this be used for some sort of fractal version of machine learning. Like the old fractal mountain landscape generator.

My idea is to see if shapes could be used on the output from the weight matrices from the various layers. That is. You have some ?grid like shape that you drag connected lines on. Making the output shape look different.

Change the shape a little and see what happens. What can it be used for?

So this could be a GIMP or Photoshop like layer where you input insert non optimizable parameters. Like colors, contrast, local scaling etc. I think it can be artist preferences.

That is. Easy changeable user input. Like the mountain seed generator where you change a single number.

Job Idea – Build Your Own Tiny Business-Home In A Startup Village? Get Your Team Build-Experience And The Confidence To Build A Company?

Like a summer house village where people buy a piece of land to build a house. Why could not a startup company build their own home from scratch.

Get knowledge from house construction to physically get a build-experience. To get your company to the position you want.

So the idea is to iterate your team member experience and confidence to a perfect start.

Sustaining the Gradient Of Vegetation Growth In Any Desert Is Important. Grab The Chance. Sustain by manual watering?

I wonder.

I think the ?only way to get sustainable rain in the arid desert is to manually water those places which has gotten som rain recently. When the rain hits those places again you will have a ?secondary effect. Then a third.

The idea is that after this the weather ?network would have picked up on the vegetation changes in that place to direct more sustainable natural rain.

A Very Important Idea – Let People Invest In Farming And Get Cheaper Food?

The idea is vital for our survival.

We cant eat solar panels but we can eat food and both use the sun for energy. Therefor for our future. For our survival. I believe we need to make smart business opportunities for people. Even if the money invested per person is small.

Food costs a lot for people every day. If they could invest in food production, farming, innovation etc. Then get ?cheaper food. Then we would create a very smart resilient future.

For instance. The deserts could be green with food crops. Its going to cost money in innovation and tech but the planet is worth it. We are going to save it.

Higher Abstract Version Of Potential Differences – Network Of Voltage Maps?

If an image can represent different voltages in the pixels then its possible to have higher version potential differences. Then there should exist something like physical voltage maps.

I suspect a network model of voltages can then be useful.

From this I would not be surprised if future batteries are aligned with gravity. Like lightning when discharging.