I implemented NEAT system according to official research paper, but I make mistake somewhere.
I decided to test my system on traditional XOR network problem which requires at least one hidden neuron. My NEAT system simply not add new neuron. To be clear, sometimes, system found result (with around 2-3 neurons in hidden layers) but after really long time (thousands of generations with generation size 50 and specie count 10).
That's not good. What can possibly cause this behaviour? I think it's because innovations are lost so specie subsystem maybe not works correctly ...
EDIT:
generation size 50
mutate rate 80%
elitism
new neuron rate 3%
new connection rate 5%
no stolen babies (all species has same number of children)
specie count 10
mating between species 0%
If you are looking for a example code, you can find it on my github: https://github.com/Lucas001/NEAT-for-autonomous-car.
But, trying to answer your question, what is your rate to add a neuron?
You can also look for help in a book called AI Techniques for Programming. The author uses c++ code to implement NEAT algorithm, while in my code it is in JAVA.
Related
I am implementing a functional face detection algorithm in C using Viola and Jones algorithm. I'm having trouble understanding Adaboost to train a strong classifier.
I can detect all 5 basic haar-features in a single image (162336 in a 24x24 image) I'm pretty sure this is good and working, and my algorithm outputs and array containing all the features sorted.
Then, I started working on Adaboost and here's what I understand. We create a weak classifier (slightly better than random) and we make a linear combination of many weak classifier (approx 200) to get a strong classifier.
What I don't understand is how to create this weak classifier. From what I read online:
Normalize the weights of our training examples (first round 1 by default)
Then get a feature (here's one of my problem, do I have to process each feature of each training example ? (162336 * number of examples) that would be a lot of computing power no ? )
"Apply" this feature to each image to get an optimal treshold and toggle (here's my main problem, I don't understand what "apply" means here, compare it with each feature of the image ? I really don't see what I have to do with it. Then, I don't understand what is the treshold and the toggle and that's where i'm looking for help)
Then many more other things to do
I'm really looking forward your help to make me understand this!
Should have answered my own question faster, but I've forgot about it. It was a project for my computer science school so I can provide answers.
Adaboost is in fact fairly simple when you understand it.
First you need to detect features inside every images in your base (we used 4000 images to have a large set) you can store them if you have enough memory or process them when you need them in you program. For 4000 images with 5 haar features inside we used more than 16Gb of RAM (Code was written in c, but no memory leak, it was arrays of double)
The training algorithm assign a weight to an image. That weight represent the difficulty for the algorithm to make a good prediction (face or no face).
Your training algorithm will be composed of rounds (200 rounds is fine to have 90%+ of good prediction).
At the first round every image possess the same weight because the algorithm never worked on them.
Here is how a round goes:
Find the best haar feature among X (for each type) in each image. To do this, compare each feature to the same one (same type, dimension and position) on every image and see if it is a good or bad prediction. The feature with the best prediction inside the X features is the best one, keep it stored.You will find 5 best features per images because there is 5 type combine them in a single struct and it is your weak classifier
Calculate the weighted error of the classifier. The weighted error is the error of the weak classifier applied to each image while taking in account the weight assigned to each image. In later rounds, the image with a bigger weight (the algorithm made lot of mistakes about this image) will be taken much more into account.
Add the weak classifier to a strong classifier (which is an array of weak classifiers) and its alpha. The alpha is the performance of the weak classifier and is determined with the weighted error. With the alpha, weak classifier which were made at later stage of the algorithm when the training is harder will have more weight in the final prediction of the strong classifier.
Update the weight of each image according to the prediction of the weak classifier you just created. If the classifier is right the weight goes down otherwise it goes up.
At the end of the 200 rounds of training, you will possess a strong classifier composed of 200 weak classifier. To make a prediction about a single image, apply each weak classifier to the image and the majority wins.
I voluntarily simplified the explanation but the majority is here. For more informations look here, it really helped me during my project: http://www.ipol.im/pub/art/2014/104/article.pdf
I suggest every person interested in AI and optimisation to work on a project like that. As a student it made me really interested in AI and made me think a lot about optimisation, which I never did before.
My professor asked my class to make a neural network to try to predict if a breast cancer is benign or malignant. To do this I'm using the Breast Cancer Wisconsin (Diagnostic) Data Set.
As a tip for doing this my professor said not all 30 atributes needs to be used as an input (there are 32, but the first 2 are the ID and Diagnosis), what I want to ask is: How am I supposed to take those 30 inputs (that would create like 100+ weights depending on how many neurons I would use) and get them into a lesser number?
I've already found how to "prune" a neural net, but I don't think that's what I want. I'm not trying to eliminate unnecessary neurons, but to shrink the input itself.
PS: Sorry for any english errors, it's not my native language.
That is a question that is being under research right now. It is called feature selection and there are some techniques already. One is Principal Componetns Analysis (PCA) that reduces the dimensionality of your dataset taking those feature that keeps the most variance. Another thing you can do is to see if there are highly corelated variables. If two inputs are highly correlated may mean that they carry almost the same information so it may be remove without worsen much the performance of your classifier. As a third technique you could use is deep-learning which is a technique that tries to learn the features that will later be used to feed your trainer. More info about deep learning and PCA can be found here http://deeplearning.stanford.edu/wiki/index.php/Main_Page
This problem is called feature selection. It is mostly the same for neural networks as for other classifiers. You could prune your dataset while retaining the most variance using PCA. To go further, you could use a greedy approach and evaluate your features one by one by training and testing your network with each feature excluded in turn.
There is a technique for feature selection using just neural networks
Split your dataset into three groups:
Training data used for supervised training
Validation data used to verify that the neural network is able to generalize
Accuracy testing used to test which of the features are required
The steps:
Train a network on your training and validation set, just like you would normally do.
Test the accuracy of the network with the third dataset.
Locate the varible which yields the smallest drop in the accuracy test above when dropped (dropped meaning always feeding a zero as the input signal )
Retrain your network with the new selection of features
Keep doing this either to the network fails to be trained or there is just one variable left.
Here is a paper on the technique
I have to develop a personality/job suitability online test for an HR department. Basically, users will answer questions, on a scale of 0-10 for example, and after say 50 questions, I want to translate that to a rating in 5 different personality/ job suitability characteristics.
I don't have any real data to start with, so first, is it even worth it to use a recommendation engine like MyMediaLite (github). How many samples will I need to train it to a decent performance?
I previously built a training course recommender, by simply doing and hand-weighted sum where each question increased the weight of several courses that were related to that question. It was an expert system, built like a feed-forward neural network, where I personally tuned all the weights based on my knowledge of the questions and the courses' content.
I would like to this time around use a recommender system, but I'm wondering how many times I would have to take the 50 question test, and then assign the results manually. would 100 examples do? that could be possible. 1000 would be too long. How can I know ahead of time?
Though useless, I want to say this is not possible to give a definite number. You should focus on learning curve when adding new samples.
You can process the samples by hand and engine on parallel, and compare the result given by both. Once the measurement e.g. recall and precision of the result given by engine achieve your expectation, then you get enough samples.
Hope this helpful!
I remember when I was in college we went over some problem where there was a smart agent that was on a grid of squares and it had to clean the squares. It was awarded points for cleaning. It also was deducted points for moving. It had to refuel every now and then and at the end it got a final score based on how many squares on the grid were dirty or clean.
I'm trying to study that problem since it was very interesting when I saw it in college, however I cannot find anything on wikipedia or anywhere online. Is there a specific name for that problem that you know about? Or maybe it was just something my teacher came up with for the class.
I'm searching for AI cleaning agent and similar things, but I don't find anything. I don't know, I'm thinking maybe it has some other name.
If you know where I can find more information about this problem I would appreciate it. Thanks.
Perhaps a "stigmergy" approach is closely related to your problem. There is a starting point here, and you can find something by searching for "dead ants" and "robots" on google scholar.
Basically: instead of modelling a precise strategy you work toward a probabilistic approach. Ants (probably) collect their deads by piling up according to a simple rule such as "if there is a pile of dead ants there, I bring this corpse hither; otherwise, I'll make a new pile". You can start by simplifying your 'cleaning' situation with that, and see where you go.
Also, I think (another?) suitable approach could be modelled with a Genetic Algorithm using a carefully chosen combination of fitness functions such as:
the end number of 'clean' tiles
the number of steps made by the robot
of course if the robots 'dies' out of starvation it automatically removes itself from the gene pool, a-la darwin awards :)
You could start by modelling a very, very simple genotype that will be 'computed' into a behaviour. Consider using a simple GA such as this one by Inman Harvey, then to each gene assign either a part of the strategy, or a complete behaviour. E.g.: if gene A is turned to 1 then the robot will try to wander randomly; if gene B is also turned to 1, then it will give priority to self-charging unless there are dirty tiles at distance X. Or use floats and model probability. Your mileage may vary but I can assure it will be fun :)
The problem is reminiscent of Shakey, although there's cleaning involved (which is like the Roomba -- a device that can also be programmed to perform these very tasks).
If the "problem space" (or room) is small enough, you can solve for an optimal solution using a simple A*-based search, but likely it won't be, since that won't leave for very interesting problems.
The machine learning approach suggested here using genetic algorithms is an interesting approach. Given the problem domain you would only have one "rule" (a move-to action, since clean could be eliminated by implicitly cleaning any square you move to that is dirty) so your learner would essentially be learning how to move around an environment. The problem there would be to build a learner that would be adaptable to any given floor plan, instead of just becoming proficient at cleaning a very specific space.
Whatever approach you have, I'd also consider doing a further meta-reasoning step if the problem sets are big enough, and use a partition approach to divide the floor up into separate areas and then conquering them one at a time.
Can you use techniques to create data to use "offline"? In that case, I'd even consider creating a "database" of optimal routes to take to clean certain floor spaces (1x1 up to, say, 5x5) that include all possible start and end squares. This is similar to "endgame databases" that game AIs use to effectively "solve" games once they reach a certain depth (c.f. Chinook).
This problem reminds me of this. A similar problem is briefly mentioned in the book Complexity as an example of a genetic algorithm. These versions are simplified though, they don't take into account fuel consumption.
Just wondering, since we've reached 1 teraflop per PC, yet we are still not able to model an insect's brain.
Has anyone seen a decent implementation of a self-learning, self-developing neural network?
I saw an interesting experiment mapping the physical neural layout of a rat's brain to a digital neural network with weighting modelled on the neuron chemistry of each component taken using MRI and others. Quite interesting. (new scientist or Focus, 2 issues ago?)
IBM Blue Brain comes to mind
http://news.bbc.co.uk/1/hi/sci/tech/8012496.stm
The problem is computation power as you rightly point out. But for a sequence of stimuli to a neural network the range of calculations tends to be exponential as that stimuli encounters deeper nested nodes. Any complex weighting algorithm means that time spent at each node can get expensive. Domain specific neural-maps tend to be quicker because they are specialized. Brains in mammals have many general paths, making it harder to teach them, and for a computer to model a real mammal brain in a given space/time.
Real brains also have tons of cross-talk like static (some people think this is where creativity or original thought stems from). Brains also don't learn using 'direct' stimulus/reward ... they use past experience of non-related matter to create their own learning. Recreating the neurons is one thing in a computational space, creating an accurate learning is another. Never-mind the dopamine (octopamine in insects) and other neurological chemicals.
imagine giving a digital brain LSD or anti-depressants. As a real simulation. Awesome. That would be a complex simulation I suspect.
I think you're kind of making the assumption that our idea of how neural networks work is a good model for the brain at a large-scale level; I'm not sure that is a good assumption. Hell, not too many years ago we didn't think the glial cells were important to mental functions, and it was the idea for a long time that there is no neurogenesis after the brain matures.
On the other hand, neural networks do seem to handle some apparently complex functions pretty well.
So, here's a little puzzle question for you: how many teraflops or petaflops do you think a human brain's computation represents?
Jeff Hawkins would say that a neural net is a poor approximation of a brain. His "On Intelligence" is a terrific read.
Yup: OpenCog is working on it.
It's the structure. Even if we had computers today with the same or higher performance than a human brain (there are different predictions when we'll get there, but there are still a few years to go), we still need to program it. And while we know a lot of the brain today, there are still many, many more things we do not know. And these aren't just details, but large areas that are not understood at all.
Focusing only on the Tera-/Peta-FLOPS is like looking only at megapixels with digital cameras: it focuses on only one value when there are many factors involved (and there are a few more of those in a brain than in a camera). I also believe that many of the estimates just how many FLOPS would be needed to simulate a brain are way off - but that's a different discussion altogether.
Just wondering, we've reached 1 teraflop per PC, and we are still not able to model an insect's brain. has anyone seen a decent implementation of a self-learning self-developing neural network?
We can already model brains. The question these days, is how fast, and how accurate.
In the beginning, there was effort expended on trying to find the most abstract representation of neurons with the least amount of physical properties needed.
This led to the invention of the perceptron at Cornell University, which is a very simple model indeed. In fact, it may have been too simple, as the famous MIT AI professor, Marvin Minsky, wrote a paper which mistakenly concluded that it would be impossible for this type of model to learn XOR (a basic logic gate that could be emulated by every computer we have today). Unfortunately, his paper plunged neural network research into the dark ages for at least 10 years.
While probably not as impressive as many would like, there are learning networks that are already in existence that can do visual and speech learning and recognition.
And even though we have faster CPUs, it is still not the same as a neuron. Neurons in our brain are, at the very least, parallel adder units. So imagine 100 billion simulated human neurons, adding each second, sending their outputs to 100 trillion connections with a "clock" of about 20hz. The amount of computation going on here far exceeds the petaflops of processing power we have, especially when our cpus are mostly serial instead of parallel.
In 2007, they simulated the equivalent of a half mouse brain for 10 seconds at half the actual speed: http://news.bbc.co.uk/1/hi/technology/6600965.stm
There is a worm named C. Elegance and its anatomy is completely know to us. Every cell is mapped out and every neuron is well studied. This worm has an interesting property by birth and that is it follows or grow towards only those temperature regions in which it was born. Here is link to the paper. This paper has implementation of the property with neuronal model. And there are some students who have built robot that only follows dark regions in the region having different shades of light, using this neuronal model. This work could have been done using other methods as well but this method is more noise resilient as proved by paper to which I have given link above.