I was wondering why pytorch did not work on my AMD x4 computer. It seems it is too old. Does not support SSE4.
So I tested with success the Intel Software Development Emulator with pytorch and cuda enabled.
Yes it worked. Using only the CPU took more time than I would like to wait. But boy using the gpu. The GPU went up to 100% and the calculation took just a second. Way fast.
So with a CUDA enabled graphics card you can run pytorch on an old cpu.
Idea – Only emulate whats missing
This gave an idea. Is it possible to make a similar emulator only for sse4 and future functions? So the emulator does not emulate everything. Just what is missing. Otherwise I think its slower than it have to be. Then we could have that running by default running in Linux.
I checked the wikipedia summation of a fuel cell. It converts chemical energy to electricity by oxidation.
So my idea is. Can you use a solid material instead hydrogen gas?
I wonder. Could there exist a reaction between silicon and oxygen if you just slam some oxygen molecules into the silicon. Is there a efficient way to create a reaction? Could this be a solvable problem. One of many to a sustainable energy future.
So the reason for the idea is that I think we need abundant materials like silicon for devices like fuel cells.
Life is not necessary for the early universe to evolve. From the talk about machine learning. All you need are algorithms.
So my speculation is that the universe starts with an algorithm. This algorithm is then responsible for fine tuning the universe towards supporting life.
As a imagination exercise.
What would happen if an electron absorbs light to an excited state in a another environment similar to that of the atom. Would the electron emit light. Could you create a force?
To get a force you need something like two charges. Two electrons maybe. So if you have something like gravity that produces a force then there should I guess be a ?temporary electron emergence caused by the gravity.
Maybe the speed of gravity being close to speed of light results in a ?temporary electron some short time after the gravity has hit the atom. Then it would result in a attracting force. An electron repelling from behind the atom.
If you rewrite a physical law as a 100% rule then a valid guess is that there are probabilistic rules.
So maybe there exists a motion law with efficiency. That is. For every energy unit in you get a chance of forward force or generation of light. Here it could be that you get 10% force and the rest as light. Maybe light generation in the atom has some odd states that only happen at a low probability. Which then generates a force.
Maybe an atom can generate a new ?temporary electron which it can use to absorb energy as motion if the energy transformation options are limited. Then from electron to ?temporary electron repulsive forces you can get motion.
Then it should be possible to create a probabilistic drive.
Part of this idea. The predicted electron generation is due to the universe being optimized not to harm life. So from this I guess that the universe is also optimized towards drinkable water. So here you got another smart power sentence to build ideas from.
( A guess of mine is that atoms can under ?very cold temperatures generate an electron cloud. Giving it some above zero kelvin temperatures. To maintain functionality. )
Since you can move a tiny house on wheels. You are better off finding a spot with low rent.
So with a low monthly cost. I wonder if a livelihood for some previously homeless people could be to cut hair. This in Tiny House furnished as a hair cutting space.
Also. A second idea I thought of was a tiny house for making pizza. The thing I wondered about was if it could use drones for delivering the food.
But this gave me another idea that retail stores could perhaps use drones to compete with online stores.
Often my iterations stall. The error or loss is too high to say the model has converged. I wonder if this a general problem.
So I will test if you can use the update history of a iteration and feed that to another network to guess the next update.
With this I have.
Update0 Input -> Network -> Update1 target
Update1 Input -> Network -> Update2 target
Update2 Input -> Network -> Update3 target
Update3 Input -> Network -> Update4 target
So when it stalls I will see if I can replace some update with a vector I got from the Network instead. Something like this.
What do we do if there is no central school to go to. No university or college.
Come to think of it. I think this is the case for many around the globe. So if we build Education resilience. I think we could use this knowledge for this challenge.
One ultra intelligent solution to this problem is to teach everyone how to teach. Teaching should be part of the curriculum for every student.
A quick idea.
Besides housing needs. Particularly in the winter. Entrepreneurship training that leads to even the smallest livelihood is good knowledge and knowledge is something that you don’t loose. So regardless if they stay or return home. They can build a business here or there. Train others in the same skills.
I believe this is a sustainable way forward. Include everyone.
It could be that there is an error in their thinking about fusion. Using a large ?tokamak device and circulating energy at the same spot is wrong!
I think safe fusion starts with ultra small devices. No need to create a device which you feed energy into the grid with right away.
Since fusion is a difficult process the universe have had to be optimize to achieve it. Then from looking at the sun. I guess that you need a complex heat treatment before fusion.
A quick idea or rather a question really.
I often run into problems my machine learning skills cant handle. So I’m wondering if I go from problem to failed solution to quickly. Maybe there exist a way to transform my problem into a simpler one using machine learning.
Hmm. I think I can pinch the model so that it produces succesive ?simplifications.
If physics and chemistry is connected why is this so? Since you can ask a smart network object a why question. The connection ?should be that there exist a network problem that needs both physics and chemistry.
One such network problem is life.
Maybe since a machine learning network function can obtain ?any function. A network problem can assume any problem.
Going from different programming languages I do make some syntax errors. From this I will test if I can teach a machine learning network to correct my most common mistakes. Perhaps not correct but to prompt me a label with an example. Come to think of it. Maybe I could have a chat bot writing some hilarious jokes while I’m coding and getting errors.
So taking this further. Would it be possible to create an alternative to Google. I wonder if its possible to write the type of problem your trying to solve in a #comment. Containing the words you use if you search it on Google. Then you could let a program go through the #comments and suggest good examples. In some pretty fashion.
I was thinking. Why ”compress” information into a small number of weight parameters and also perform value prediction with them. This many jobs for the weights I think just makes it more difficult.
So I thought. Could you split the process of training. I mean. A time series could perhaps be split into several data series. Some sort of extra parameters.
What if the weights are not constants but stored in a time series themselves. Now that we have thought of a large storage space for parameters the training of them should also be made easier. What if they could contain errors. A good enough solution is better than no solution. So when when training the large parameter list we should only train those parameters that make sense. Otherwise it would take too long.
Even if you start helping one group. The probability increases for others to get help as the momentum and knowledge builds up in total. This because challenges are similar in some sense.
Take refugee housing. Going from tents to housing units is a big improvement. Similar problems exist for the homeless. Here you also have tiny houses from living on the street.
The next challenge is income. Jobs and entrepreneurship. Here you could help the largest group by thinking big. Because the need to innovate you build up knowledge. As this process progresses you apply the knowledge to every group.
So I thought. How could retail compete with online stores. What would I like.
I like going to theme parks so why not combine tiny house stores in a little village like simple theme park or ?Circus.
The idea here is to see if there is such a thing as a continuum method.
Let say you train a hidden layer network from X as input and X as output. This should converge pretty quickly. So I thought in a time series. Like the weather temperature. The overall the temperature series over a year have a characteristic look. Then if I start the network with its weights. Such that the network gives a temperature characteristic graph. Then I hope if I change the input data a little the weights only need to change a little.
It might be that the weight matrices need to be smooth and not so fragile.
The idea with this was to calculate predictions faster.
example of smooth weight matrices,syn0 and syn1. I used in a feedforward network with max(ErrorList) error function with fmin_slsqp optimation.
syn0 = syn0.reshape(32,32)
syn1 = syn1.reshape(16,16)
syn0 = sp.ndimage.filters.gaussian_filter(syn0, sigma=[1,1], mode=’constant’)
syn1 = sp.ndimage.filters.gaussian_filter(syn1, sigma=[1,1], mode=’constant’)
errorList[i] = np.sum((y – l2)**2)
maxError = np.max(errorList)
Anim showing succesive iterations of syn0
I cant say I know much about fuel cells. So when I looked at a fuel cell I wondered. Since the membranes are stacked. Have the system a similar problem of efficiency as stacked peltier elements?
If so then it might be an idea to use many small thin fuel cell units. Look at https://www.greenoptimistic.com/smallest-hydrogen-fuel-cell-20090113/ Something like that.
What if you can make a small fuel cell unit very efficient and scale that thousands of times. Could this large cluster power something more than electronics?
I mean to charge the electric car battery on the go or at stops.
Do not accept terrible conclusions. Here you can suggest several good ideas instead.
If a problem is too difficult. Then innovate ideas around other problems related to the discussed problem. It could be that many problems are related but it does not appear so at first.
I believe its harder to get stuck in bad thinking if you activate your brain. That is. To understand a new innovative idea. This instead of letting your brain feed on circulating negative information.
Consider 1 = 0.999…
Inspiration taken from this statement give me.
To think logical is the feeling of succeeding with a shortcut.
The information in the numbers according to the scientific notation is m * 10^n. So I had the idea that I in machine learning problems separate this information into two data columns.
I think this could be some kind of method but I haven’t fully tested it yet.
Got the exact same result after testing. Kinda strange.
How do you prevent beatings, property damage and improve student well being? Could there exist many innovations we have not thought of yet?
Prevention through connected after school activities?
Machine learning is the new so why not add machine learning algorithms to these office suits.
So with these you just mark training data columns and target value columns. Choose the model and fit. Then predict data from the model.