Why does the brain use neural networks for solving different tasks? I wonder. Are there more advantages to this. Other than that we need to learn first.
From this I wonder what would happen if you let a machine learning setup learn different mathematical operators. Like addition, subtraction, multiplication and division. Using several different neural network setups.
The simple addition should then be expanded into a larger matrix of learned weights. Similarly for other functions or operators.
What I then suspect is that you can record what happens in the neural network setup when doing calculations. This recording I imagine is similar to the feeling we as humans get when we do calculations.
From this idea you get something of a new field. Feeling based mathematics. Here you can probably solve different tasks better with feelings as help data.
The idea is very simple and I think I should have used it all along.
Whenever you have an idea. Just launch your text editor and write the idea. The images from which you got inspiration will be on the screenshot you take of the whole screen.
Simple and very effective.
Further I wonder if you can take a screencast while creating music. That way you can store the feeling you had. Quickly capture everything that happened that day and go back and improve. Find out what others thought was good that one perticular day and so on.
What I would need in Linux is a transparent drawing window. On this I could draw red arrows and underlines for any program I like.
The idea is that a program or a smart manual can collect information from your coding. Suggest error preventing information, related information, enhancements and more.
This while you type in a window next to this information window.
Newly installed Ubuntu MATE with Focuswriter and the Ambient Noise software for the forest sound.
I think one can simplify this idea further. Let the programmer himself write the information/tip piece that is presented to him. So here you create a database of useful tips. So when you write the programming keyword or method name. An own defined text with tips is shown in the split window.
Sometimes you want to connect two computers but you only have a web browser at your disposal. Like when running a Linux distro live USB.
I had not thought about it before but you can use use your android phone for such a simple task. Just install some WiFi file transfer program on the phone. Then connect two or more computers to that webserver on the phone through the webbrowser of each computer.
You can connect several computers to one webserver so they will be able to share files.
This works for small files since I found it a bit slow. Just use the download folder on the phone.
Every time I make a clean install of the operating system I believe I’m wasting my time selecting which software to reinstall.
So why couldn’t the Software Boutique in Ubuntu MATE have a bookmark capability. Like Firefox. It should be able to store bookmarks of programs you want to reinstall in one batch. Similar to a script.
One way I think this could work is if the boutique can handle several lists. Like preferred music lists. This way you could create a new list from many old ones. Pick different software from different lists and just install the programs you want.
You don’t need to make it cloud based. Just provide an import/export function like the bookmark function in Firefox.
This way is much faster than having to search for every program.
I think there should exist a website dedicated to helping users like myself trying to compile open source software.
Now I really want to test Fugio. But my efforts trying to compile it have failed. Thats why I believe Linux users need to have a website just for explaining how to compile various software.
“Fugio (pronounced foo-gee-oh) is an open visual programming system for building digital art and creative projects quickly, with no programming experience required.” – http://www.bigfug.com/software/fugio/
To get a new idea you might want to try this approach.
I will start with an example.
Video recording => Deterministic Video recording
Here I first added ‘deterministic’ to the invention with the requirement that it does not change it. The more words the better. Just make changes to the added words and get new inspiration.
From this rewrite I could sense that digital recordings are better because of the deterministic results.
The inspiration I get from this is that determinism is just the realization that everything can be recorded after it has happened.
Also what would an almost deterministic video be like? Could it be calculated in some parts and be recordable in others. Recordable here means that if you play it several times. Those parts will remain the same.
Then I wonder if determinism could be splitted into short time recordable parts and need to be calculated parts.
The usual way people respond to this is that all changes can be recorded. Sure but it would be very wasteful if you need to record a million times.
Then I wonder if the part that needs to be quickly calculated is information that is recognized as a new set.
I guess this fast set generation is the freewill system. It will go towards maximum complexity with the least amount of required information.
The important thing from this post is that you can classify the main word any way you like to get new inspiration.
Take a color photo. Separate the different colors and sort them. Plotting all three color channels red, green and blue. It will reveal something like the left image. The right shows all the channels mixed together and sorted.
From this I wonder. Could you with the graph on the right side create the three curves to the left. Does there exist a clustering method that will do this? Split the long vector into three different looking enough vectors. Perhaps the separation could be found somehow.
One way I think it could work is by cheating a little. By using a small scaled down version of the photo as a guide.
Such a method could be part of a new image compression system. Here you store a small scaled down version of the photo and lossy compressed enlargement information. This information is then used together with the small photo as a guide to get the big photo.
If this works I can formulate another power sentence. Everything can be improved with a guide.
The idea is simple. I speculated before about word objects. These objects are just collections of words and numbers. The words explain what you mean and the number its score of truth.
I personally think this is a x idea.
So my idea is that you can make two versions of the text. One ordinary and one with word objects and their score. Maybe you can have a tool tip or a pop-up menu on some of the important words.
If you then express yourself as detailed as you can. Then others can compare their own subjective scores with yours.
Since you can write with word objects you don’t need to choose the words so precisely. You just score it with a number or a sentence explaining the relation. This way you don’t filter the sentences to get to the perfect words.
I wonder. Should there not be a reward or any kind of mathematical gain for doing transformations. Lossless transformations in particular. I think this could be true.
One kind of transformation I think of is sorting to a index and value matrix of a matrix.
I did this for a random matrix of some size. The idea is that this transformation into a real valued matrix and two sorted matrixes of integers will better, faster or something when doing some calculation.
Since ?everything can be described by words. Like objects, inventions and more. I was wondering. Could there exist middle words. Word objects we have not yet put a name on. If so. Then the above sentence is false. There should exist inventions that have new words.
So is there a ?better way to come up with new words. Or rather is there a better way to generate new word objects. Maybe something simple and self explanatory.
I imagine a word object could be like a matrix with associated words and a score value describing the new word. These word objects would make what we mean more exact.
So an example would be this abstract70 and imaginative90 idea.
Simple Word Objects For when you do not know the middle word.
Oh I got an exciting idea today. Why use the boring Libreoffice Impress or Powerpoint. You can customize a very interesting presentation app using a game engine.
All this can be done with open source software on Linux.
I plan to use the Godot Game engine, Blender and GIMP.
The idea is kind of fresh so I will return and reedit this post after some days.
So the idea is to program a presentation application using the Godot Game engine. Since its uses 3D acceleration the presentations will look innovative.
You can create a 3D world you can interact with during the show. Or why not present a 3D car simulation for that new electric car.
With a game engine you can really make some super cool 3D presentations.
For those of us like my self. Writing is not so fluent. I forget words and sentences sometimes get too long.
So the idea is to let someone else read. Then if a syntesized speech voice could read you the text and it sounds ok. Well then you know it will be readable for everyone else. Usually you only need to shorten or rewrite sentences.
To capture audio in Linux. I used the somewhat odd solution of recording a screencast with simplescreenrecorder. Then just strip the audio from it with ffmpeg.
The idea is simple. Innovation can be I assume clustered into groups. What I imagine is that the smaller the innovations are. The more similar the innovations become. This because of the less degree of freedom.
This would mean there is a connection between biology and technological innovation. I think biomimicry tries to do this. Take innovation from biology to technology.
I imagine a similar approach. Since the innovations should be more similar at say the ?nano level. A machine learning approach should be able to learn from images from those biological ”innovations”.
The idea is then that you present a yet not finished technological innovation to the machine learning program. It could then use this as enough information for generation. Inserting probable pixels in the generated image. From analyzing the generated image. An engineer would then know what the innovation should look like from the perspective of the machine learning program. This based on what it had learned from biology.
Maybe this approach could be used for a new battery?
The idea is. With lots of data representing a signal you want to use. You might get lots of different values for the same x coordinate. That is. More than one y value for the x coordinate.
Then I thought why reduce a signal to a function of statisitical points. With some least square method for some line. I mean. We have computers capabable of taking advantage of as many data points we can sample.
Still to make some size reduction my idea is to store all the data in some kind of fourier series. That is. With multiple values y for some x coordinate. Not making some statistical reduction to get a function.
This information I imagine could be useful for a machine learning algorithm.
An example would be a camera photo. Where you will have multiple colors for some of the pixels. Not reducing the pixel color to one color. Just storing all the colors that the pixel got from the sensor. I think this could be taking advantage of in machine learning.
Since there are several statistical methods available. Some will be better than others for the particular problem. However my hope is that machine learning will find the best method possible. And since machine learning evolves its better to store all the data for future improvements.
Another idea would be to keep the value of one of the multiple values in every multivalued point. So that in a lossy fourier based compression. Each of the “choice values” makes a good contribution to the signal. By good. I mean it does not add to some ?high frequency noise.
The idea is simple. When you take photos you almost always want to take with the highest resolution. Because you can not retake the photo. Then there is a problem. Since I think the standard for the highest resolution is a 4:3 box. For my phone its 2592×1944 = 4*648 x 3*648 or a 4:3 ratio . But the screen has a resolution that is 16:9 in resolution.
Then the idea is to have help lines showing a 16:9 box in the phone camera application. You then aim so that the important stuff is in the 16:9 box. Even if the photo will be a 4:3 box.
Here the idea is to implement automatic crop and store two photos. One with the 4:3 resolution which can be printed on paper and a 16:9 resolution perfect for viewing on the TV or phone.
Perhaps the meta data about the position of the 16:9 box could be stored in the larger 4:3 photo. But when viewing on the TV or the phone it automatically crops the image to the right position.
Maybe a default position like in my screen shot could be used. Equal cut on one side as the other.
The inner box is the 16:9 box of help lines. Then the 4:3 box. The actual photo. This is all wrapped up in the phone screen.
Blender 3D which I used for this is an excellent program on Linux for testing ideas.