I'm working on a web application which will be used for classifying photos of automobiles. The users will be presented with photos of various vehicles, and will be asked to answer a series of questions about what they see. The results will be recorded to a database, averaged, and displayed.
I'm looking for algorithms to help me identify users which frequently don't vote with the group, indicating that they're probably either not paying attention to the photos, or that they're lying about what they see. I then want to exclude these users, and recalculate the results, such that I can say, with a known amount of confidence, that this particular photo shows a vehicle that is this and that.
This question goes out to all you computer science guys, where to find such algorithms or to give myself the theoretical background to design such algorithms. I'm assuming I'm going to have to learn some probability and statics, maybe some data mining. Some book recommendations would be great. Thanks!
P.S. These are multiple choice questions.
All of these are good suggestions. Thank you! I wish there was a way on stack overflow to select multiple correct answers so more of you could be acknowledged for your contributions!!
Read The Elements of Statistical Learning, it is a great compendium on data mining.
You can be interested especially in unsupervised algorithms, for example clustering. Assuming that most people do not lie, the biggest cluster is right and the rest is wrong. Mark people accordingly, then apply some bayesian statistics and you'll be done.
Of course, most data mining technologies are pretty experimentative, so don't count on that they will be always right... or even in most cases.
I believe what you described is solved using outlier/anomaly detection.
A number of techniques exist:
statistical-based methods
distance-based methods
model-based methods
I suggest you take a look at these slides from the excellent book Introduction to Data Mining
If you know what answers you are expecting why do you ask people to vote? By excluding some values you basically turn the vote in something that you like. Automobiles make different impression to different individuals. If 100 ppl loved a car then when someone comes and says that he/she doesn't like it, you exclude the vote?
But anyway, considering that you still want to do this, first of all you will need a large set o data from "trusted" voters. This will give you an idea of "good" answer and from this point you can choose the exclude threshold.
Without an initial set of data you cannot apply any algorithm because you will get false results. Consider just one vote of 100 from on a scale from 0 to 100. The second vote is "1" The you will exclude this vote because is too far away from the average.
I think a pretty simple algorithm could accomplish this for you. You could try and get fancier by calculating the standard deviations and such, but I wouldn't bother.
Here's a simple approach that should be sufficient:
For each of your users, calculate the number of questions they answered and the number of times they selected the most popular answer for the question. The users which have the lowest ratio of picking the popular answer versus total answers you can guess are providing bogus data.
You probably would not want to throw out the data from users where they've only answered a small number of questions because they likely have just disagreed on a few versus putting in bogus data.
What kind of questions are they (Yes/No, or 1 to 10?).
You may be able to get away with not discarding anything by using a mean instead of an average. With averages if there are extreme outliers in the response it could affect the average, but if you use median you may get a better answer. So for example if you had 5 answers, order them and pick the middle one.
I think what you are saying is that you are concerned that certain people are "outliers", and they are adding noise to your data, making the categorizations less reliable. So, if you have a Chevy Camaro, and most people say it is either a pony car, a muscle car, or a sports car, but you have some goofball who says it's a family sedan, you would want to minimize the impact of his vote.
One thing you could do is provide a Stack Overflow-like reputation score for users:
The more a user is "in agreement" with other users, the better his or her score would be. For a given user (User X), this could be determined by a simple calculation of what percentage of users who responded to a question chose the same category as User X, then averaging this value over all questions answered.
You may want to to multiply this value by the total number of question answered to encourage people to answer as many questions as possible. (Note: if you choose to do this, it would be equivalent to just summing the percentage agreement scores rather than averaging them.)
You could present the final reputation score to users, making sure to explain that they will be rewarded for how well their responses agree with those of other users. This will encourage people to answer more questions but also to take care in their answers.
Finally, you could calculate a certainty score for a given categorization by adding up the total reputation score of all people who chose a given category.
Some of these ideas may need some refinement, especially since I don't know your exact situation. Certainly, if people can see what other people chose before they vote, it would be way too easy to game the system.
If you were to collect votes like "on a scale from 1 to 10, how would you rate this car", you could probably use simple average and standard deviation: the smaller the standard deviation, the more unanimous the general consensus is among your voters, and you can flag users who are e.g. 3 standard devs from the average.
For multiple choice, you need to be more careful. Simply discarding all but the most-voted option will do nothing but disgruntle the voters. You need to establish a measure of how significant the winner is w.r.t. the other options, e.g. flag users who voted for options with less than 1/3 of the winning options count.
Note that I wrote "flag users", not discard votes. If you discard votes, you can't tell how confident you are about the result ("91% voted this to be a Ford Mustang"). If a user has more than a certain percentage of his votes flagged - well, that's up to you.
Your trickiest problem, however, will probably be to collect sufficient votes. Depending on how easy the multiple choice problem is, you probably need several times the number of options as votes, per photo. Otherwise the statistics are meaningless.
Related
I'm trying to understand how random forest works in plain English instead of mathematics. Can anybody give me a really simple explanation of how this algorithm works?
As far as I understand, we feed the features and labels without telling the algorithm which feature should be classified as which label? As I used to do Naive Bayes which is based on probability we need to tell which feature should be which label. Am I completely far off?
If I can get any very simple explanation I'd be really appreciated.
RandomForest uses a so-called bagging approach. The idea is based on the classic bias-variance trade off. Suppose that we have a set (say N) of overfitted estimators that have low bias but high cross-sample-variance. So low bias is good and we want to keep it, high variance is bad and we want to reduce it. RandomForest tries to achieve this by doing a so-called bootstraps/sub-sampling (as #Alexander mentioned, this is a combination of bootstrap sampling on both observations and features). The prediction is the average of individual estimators so the low-bias property is successfully preserved. And further by Central Limit Theorem, the variance of this sample average has a variance equal to variance of individual estimator divided by square root of N. So now, it has both low-bias and low-variance properties, and this is why RandomForest often outperforms stand-alone estimator.
Adding on to the above two answers, Since you mentioned a simple explanation. Here is a write up that I feel is the most simple way you can explain random forests.
Credits go to Edwin Chen for the simple explanation here in layman terms for random forests. Posting the same below.
Suppose you’re very indecisive, so whenever you want to watch a movie, you ask your friend Willow if she thinks you’ll like it. In order to answer, Willow first needs to figure out what movies you like, so you give her a bunch of movies and tell her whether you liked each one or not (i.e., you give her a labeled training set). Then, when you ask her if she thinks you’ll like movie X or not, she plays a 20 questions-like game with IMDB, asking questions like “Is X a romantic movie?”, “Does Johnny Depp star in X?”, and so on. She asks more informative questions first (i.e., she maximizes the information gain of each question), and gives you a yes/no answer at the end.
Thus, Willow is a decision tree for your movie preferences.
But Willow is only human, so she doesn’t always generalize your preferences very well (i.e., she overfits). In order to get more accurate recommendations, you’d like to ask a bunch of your friends and watch movie X if most of them say they think you’ll like it. That is, instead of asking only Willow, you want to ask Woody, Apple, and Cartman as well, and they vote on whether you’ll like a movie (i.e., you build an ensemble classifier, aka a forest in this case).
Now you don’t want each of your friends to do the same thing and give you the same answer, so you first give each of them slightly different data. After all, you’re not absolutely sure of your preferences yourself – you told Willow you loved Titanic, but maybe you were just happy that day because it was your birthday, so maybe some of your friends shouldn’t use the fact that you liked Titanic in making their recommendations. Or maybe you told her you loved Cinderella, but actually you really really loved it, so some of your friends should give Cinderella more weight. So instead of giving your friends the same data you gave Willow, you give them slightly perturbed versions. You don’t change your love/hate decisions, you just say you love/hate some movies a little more or less (formally, you give each of your friends a bootstrapped version of your original training data). For example, whereas you told Willow that you liked Black Swan and Harry Potter and disliked Avatar, you tell Woody that you liked Black Swan so much you watched it twice, you disliked Avatar, and don’t mention Harry Potter at all.
By using this ensemble, you hope that while each of your friends gives somewhat idiosyncratic recommendations (Willow thinks you like vampire movies more than you do, Woody thinks you like Pixar movies, and Cartman thinks you just hate everything), the errors get canceled out in the majority. Thus, your friends now form a bagged (bootstrap aggregated) forest of your movie preferences.
There’s still one problem with your data, however. While you loved both Titanic and Inception, it wasn’t because you like movies that star Leonardo DiCaprio. Maybe you liked both movies for other reasons. Thus, you don’t want your friends to all base their recommendations on whether Leo is in a movie or not. So when each friend asks IMDB a question, only a random subset of the possible questions is allowed (i.e., when you’re building a decision tree, at each node you use some randomness in selecting the attribute to split on, say by randomly selecting an attribute or by selecting an attribute from a random subset). This means your friends aren’t allowed to ask whether Leonardo DiCaprio is in the movie whenever they want. So whereas previously you injected randomness at the data level, by perturbing your movie preferences slightly, now you’re injecting randomness at the model level, by making your friends ask different questions at different times.
And so your friends now form a random forest.
I will try to give another complementary explanation with simple words.
A random forest is a collection of random decision trees (of number n_estimators in sklearn).
What you need to understand is how to build one random decision tree.
Roughly speaking, to build a random decision tree you start from a subset of your training samples. At each node you will draw randomly a subset of features (number determined by max_features in sklearn). For each of these features you will test different thresholds and see how they split your samples according to a given criterion (generally entropy or gini, criterion parameter in sklearn). Then you will keep the feature and its threshold that best split your data and record it in the node.
When the construction of the tree ends (it can be for different reasons: maximum depth is reached (max_depth in sklearn), minimum sample number is reached (min_samples_leaf in sklearn) etc.) you look at the samples in each leaf and keep the frequency of the labels.
As a result, it is like the tree gives you a partition of your training samples according to meaningful features.
As each node is built from features taken randomly, you understand that each tree built in this way will be different. This contributes to the good compromise between bias and variance, as explained by #Jianxun Li.
Then in testing mode, a test sample will go through each tree, giving you label frequencies for each tree. The most represented label is generally the final classification result.
I am interested in writing a twenty questions algorithm similar to what akinator and, to a lesser extent, 20q.net uses. The latter seems to focus more on objects, explicitly telling you not to think of persons or places. One could say that akinator is more general, allowing you to think of literally anything, including abstractions such as "my brother".
The problem with this is that I don't know what algorithm these sites use, but from what I read they seem to be using a probabilistic approach in which questions are given a certain fitness based on how many times they have lead to correct guesses. This SO question presents several techniques, but rather vaguely, and I would be interested in more details.
So, what could be an accurate and efficient algorithm for playing twenty questions?
I am interested in details regarding:
What question to ask next.
How to make the best guess at the end of the 20 questions.
How to insert a new object and a new question into the database.
How to query (1, 2) and update (3) the database efficiently.
I realize this may not be easy and I'm not asking for code or a 2000 words presentation. Just a few sentences about each operation and the underlying data structures should be enough to get me started.
Update, 10+ years later
I'm now hosting a (WIP, but functional) implementation here: https://twentyq.evobyte.org/ with the code here: https://github.com/evobyte-apps/open-20-questions. It's based on the same rough idea listed below.
Well, over three years later, I did it (although I didn't work full time on it). I hosted a crude implementation at http://twentyquestions.azurewebsites.net/ if anyone is interested (please don't teach it too much wrong stuff yet!).
It wasn't that hard, but I would say it's the non-intuitive kind of not hard that you don't immediately think of. My methods include some trivial fitness-based ranking, ideas from reinforcement learning and a round-robin method of scheduling new questions to be asked. All of this is implemented on a normalized relational database.
My basic ideas follow. If anyone is interested, I will share code as well, just contact me. I plan on making it open source eventually, but once I have done a bit more testing and reworking. So, my ideas:
an Entities table that holds the characters and objects played;
a Questions table that holds the questions, which are also submitted by users;
an EntityQuestions table holds entity-question relations. This holds the number of times each answer was given for each question in relation to each entity (well, those for which the question was asked for anyway). It also has a Fitness field, used for ranking questions from "more general" down to "more specific";
a GameEntities table is used for ranking the entities according to the answers given so far for each on-going game. An answer of A to a question Q pushes up all the entities for which the majority answer to question Q is A;
The first question asked is picked from those with the highest sum of fitnesses across the EntityQuestions table;
Each next question is picked from those with the highest fitness associated with the currently top entries in the GameEntities table. Questions for which the expected answer is Yes are favored even before the fitness, because these have more chances of consolidating the current top ranked entity;
If the system is quite sure of the answer even before all 20 questions have been asked, it will start asking questions not associated with its answer, so as to learn more about that entity. This is done in a round-robin fashion from the global questions pool right now. Discussion: is round-robin fine, or should it be fully random?
Premature answers are also given under certain conditions and probabilities;
Guesses are given based on the rankings in GameEntities. This allows the system to account for lies as well, because it never eliminates any possibility, just decreases its likeliness of being the answer;
After each game, the fitness and answers statistics are updated accordingly: fitness values for entity-question associations decrease if the game was lost, and increase otherwise.
I can provide more details if anyone is interested. I am also open to collaborating on improving the algorithms and implementation.
This is a very interesting question. Unfortunately I don't have a full answer, let me just write down the ideas I could come up with in 10 minutes:
If you are able to halve the set of available answers on each question, you can distinguish between 2^20 ~ 1 million "objects". Your set is probably going to be larger, so it's right to assume that sometimes you have to make a guess.
You want to maximize utility. Some objects are chosen more often than others. If you want to make good guesses you have to take into consideration the weight of each object (= the probability of that object being picked) when creating the tree.
If you trust a little bit of your users you can gain knowledge based on their answers. This also means that you cannot use a static tree to ask questions because then you'll get the answers for the same questions.. and you'll learn nothing new if you encounter with the same object.
If a simple question is not able to divide the set to two halves, you could combine them to get better results: eg: "is the object green or blue?". "green or has a round shape?"
I am trying try to write a python implementation using a naïve Bayesian network for learning and minimizing the expected entropy after the question has been answered as criterium for selecting a question (with an epsilon chance of selecting a random question in order to learn more about that question), following the ideas in http://lists.canonical.org/pipermail/kragen-tol/2010-March/000912.html. I have put what I got so far on github.
Preferably choose questions with low remaining entropy expectation. (For putting together something quickly, I stole from ε-greedy multi-armed bandit learning and use: With probability 1–ε: Ask the question with the lowest remaining entropy expectation. With probability ε: Ask any random question. However, this approach seems far from optimal.)
Since my approach is a Bayesian network, I obtain the probabilities of the objects and can ask for the most probable object.
A new object is added as new column to the probabilities matrix, with low a priori probability and the answers to the questions as given if given or as guessed by the Bayes network if not given. (I expect that this second part would work much better if I would add Bayes network structure learning instead of just using naive Bayes.)
Similarly, a new question is a new row in the matrix. If it comes from user input, probably only very few answer probabilities are known, the rest needs to be guessed. (In general, if you can get objects by asking for properties, you can obtain properties by asking if given objects have them or not, and the transformation between these is essentially Bayes' theorem and breaks down to transposition in the easiest case. The guessing quality should improve again once the network has an appropriate structure.)
(This is a problem, since I calculate lots of probabilities. My goal is to do it using database-oriented sparse tensor calculations optimized for working with weighted directed acyclic graphs.)
It would be interesting to see how good a decision tree based algorithm would serve you. The trick here is purely in the learning/sorting of the tree. I'd like to note that this is stuff I remember from AI class and student work in the AI working group and should be taken with a semi-large grain (or nugget) of salt.
To answer the questions:
You just walk the tree :)
This is a big downside of decision trees. You'd only have one guess that can be attached to the end nodes of the tree at depth 20 (or earlier, if the tree is still sparse).
There are whole books dedicated to this topic. As far as I remember from AI class you try minimize entropy at all times, so you want to ask questions that ideally divide the set of remaining objects into two sets of equal size. I'm afraid you'd have to look this up in AI books.
Decision trees are highly efficient during the query phase, as you literally walk the tree and follow the 'yes' or 'no' branch at each node. Update efficiency depends on the learning algorithm applied. You might be able to do this offline as in a nightly batched update or something like that.
What's a good algorithm for suggesting things that someone might like based on their previous choices? (e.g. as popularised by Amazon to suggest books, and used in services like iRate Radio or YAPE where you get suggestions by rating items)
Simple and straightforward (order cart):
Keep a list of transactions in terms of what items were ordered together. For instance when someone buys a camcorder on Amazon, they also buy media for recording at the same time.
When deciding what is "suggested" on a given product page, look at all the orders where that product was ordered, count all the other items purchased at the same time, and then display the top 5 items that were most frequently purchased at the same time.
You can expand it from there based not only on orders, but what people searched for in sequence on the website, etc.
In terms of a rating system (ie, movie ratings):
It becomes more difficult when you throw in ratings. Rather than a discrete basket of items one has purchased, you have a customer history of item ratings.
At that point you're looking at data mining, and the complexity is tremendous.
A simple algorithm, though, isn't far from the above, but it takes a different form. Take the customer's highest rated items, and the lowest rated items, and find other customers with similar highest rated and lowest rated lists. You want to match them with others that have similar extreme likes and dislikes - if you focus on likes only, then when you suggest something they hate, you'll have given them a bad experience. In suggestions systems you always want to err on the side of "lukewarm" experience rather than "hate" because one bad experience will sour them from using the suggestions.
Suggest items in other's highest lists to the customer.
Consider looking at "What is a Good Recommendation Algorithm?" and its discussion on Hacker News.
There isn't a definitive answer and it's highly unlikely there is a standard algorithm for that.
How you do that heavily depends on the kind of data you want to relate and how it is organized. It depends on how you define "related" in the scope of your application.
Often the simplest thought produces good results. In the case of books, if you have a database with several attributes per book entry (say author, date, genre etc.) you can simply chose to suggest a random set of books from the same author, the same genre, similar titles and others like that.
However, you can always try more complicated stuff. Keeping a record of other users that required this "product" and suggest other "products" those users required in the past (product can be anything from a book, to a song to anything you can imagine). Something that most major sites that have a suggest function do (although they probably take in a lot of information, from product attributes to demographics, to best serve the client).
Or you can even resort to so called AI; neural networks can be constructed that take in all those are attributes of the product and try (based on previous observations) to relate it to others and update themselves.
A mix of any of those cases might work for you.
I would personally recommend thinking about how you want the algorithm to work and how to suggest related "products". Then, you can explore all the options: from simple to complicated and balance your needs.
Recommended products algorithms are huge business now a days. NetFlix for one is offering 100,000 for only minor increases in the accuracy of their algorithm.
As you have deduced by the answers so far, and indeed as you suggest, this is a large and complex topic. I can't give you an answer, at least nothing that hasn't already been said, but I an point you to a couple of excellent books on the topic:
Programming CI:
http://oreilly.com/catalog/9780596529321/
is a fairly gentle introduction with
samples in Python.
CI In Action:
http://www.manning.com/alag looks a
bit more in depth (but I've only just
read the first chapter or 2) and has
examples in Java.
I think doing a Google on Least Mean Square Regression (or something like that) might give you something to chew on.
I think most of the useful advice has already been suggested but I thought I'll just put in how I would go about it, just thinking though, since I haven't done anything like this.
First I Would find where in the application I will sample the data to be used, so If I have a store it will probably in the check out. Then I would save a relation between each item in the checkout cart.
now if a user goes to an items page I can count the number of relations from other items and pick for example the 5 items with the highest number of relation to the selected item.
I know its simple, and there are probably better ways.
But I hope it helps
Market basket analysis is the field of study you're looking for:
Microsoft offers two suitable algorithms with their Analysis server:
Microsoft Association Algorithm Microsoft Decision Trees Algorithm
Check out this msdn article for suggestions on how best to use Analysis Services to solve this problem.
link text
there is a recommendation platform created by amazon called Certona, you may find this useful, it is used by companies such as B&Q and Screwfix find more information at www.certona.com/
I'd like to rank a collection of landscape images by making a game whereby site visitors can rate them, in order to find out which images people find the most appealing.
What would be a good method of doing that?
Hot-or-Not style? I.e. show a single image, ask the user to rank it from 1-10. As I see it, this allows me to average the scores, and I would just need to ensure that I get an even distribution of votes across all the images. Fairly simple to implement.
Pick A-or-B? I.e. show two images, ask user to pick the better one. This is appealing as there is no numerical ranking, it's just a comparison. But how would I implement it? My first thought was to do it as a quicksort, with the comparison operations being provided by humans, and once completed, simply repeat the sort ad-infinitum.
How would you do it?
If you need numbers, I'm talking about one million images, on a site with 20,000 daily visits. I'd imagine a small proportion might play the game, for the sake of argument, lets say I can generate 2,000 human sort operations a day! It's a non-profit website, and the terminally curious will find it through my profile :)
As others have said, ranking 1-10 does not work that well because people have different levels.
The problem with the Pick A-or-B method is that its not guaranteed for the system to be transitive (A can beat B, but B beats C, and C beats A). Having nontransitive comparison operators breaks sorting algorithms. With quicksort, against this example, the letters not chosen as the pivot will be incorrectly ranked against each other.
At any given time, you want an absolute ranking of all the pictures (even if some/all of them are tied). You also want your ranking not to change unless someone votes.
I would use the Pick A-or-B (or tie) method, but determine ranking similar to the Elo ratings system which is used for rankings in 2 player games (originally chess):
The Elo player-rating
system compares players’ match records
against their opponents’ match records
and determines the probability of the
player winning the matchup. This
probability factor determines how many
points a players’ rating goes up or
down based on the results of each
match. When a player defeats an
opponent with a higher rating, the
player’s rating goes up more than if
he or she defeated a player with a
lower rating (since players should
defeat opponents who have lower
ratings).
The Elo System:
All new players start out with a base rating of 1600
WinProbability = 1/(10^(( Opponent’s Current Rating–Player’s Current Rating)/400) + 1)
ScoringPt = 1 point if they win the match, 0 if they lose, and 0.5 for a draw.
Player’s New Rating = Player’s Old Rating + (K-Value * (ScoringPt–Player’s Win Probability))
Replace "players" with pictures and you have a simple way of adjusting both pictures' rating based on a formula. You can then perform a ranking using those numeric scores. (K-Value here is the "Level" of the tournament. It's 8-16 for small local tournaments and 24-32 for larger invitationals/regionals. You can just use a constant like 20).
With this method, you only need to keep one number for each picture which is a lot less memory intensive than keeping the individual ranks of each picture to each other picture.
EDIT: Added a little more meat based on comments.
Most naive approaches to the problem have some serious issues. The worst is how bash.org and qdb.us displays quotes - users can vote a quote up (+1) or down (-1), and the list of best quotes is sorted by the total net score. This suffers from a horrible time bias - older quotes have accumulated huge numbers of positive votes via simple longevity even if they're only marginally humorous. This algorithm might make sense if jokes got funnier as they got older but - trust me - they don't.
There are various attempts to fix this - looking at the number of positive votes per time period, weighting more recent votes, implementing a decay system for older votes, calculating the ratio of positive to negative votes, etc. Most suffer from other flaws.
The best solution - I think - is the one that the websites The Funniest The Cutest, The Fairest, and Best Thing use - a modified Condorcet voting system:
The system gives each one a number based on, out of the things that it has faced, what percentage of them it usually beats. So each one gets the percentage score NumberOfThingsIBeat / (NumberOfThingsIBeat + NumberOfThingsThatBeatMe). Also, things are barred from the top list until they've been compared to a reasonable percentage of the set.
If there's a Condorcet winner in the set, this method will find it. Since that's unlikely, given the statistical nature, it finds the one that's the "closest" to being a Condorcet winner.
For more information on implementing such systems the Wikipedia page on Ranked Pairs should be helpful.
The algorithm requires people to compare two objects (your Pick-A-or-B option), but frankly, that's a good thing. I believe it's very well accepted in decision theory that humans are vastly better at comparing two objects than they are at abstract ranking. Millions of years of evolution make us good at picking the best apple off the tree, but terrible at deciding how closely the apple we picked hews to the true Platonic Form of appleness. (This is, by the way, why the Analytic Hierarchy Process is so nifty...but that's getting a bit off topic.)
One final point to make is that SO uses an algorithm to find the best answers which is very similar to bash.org's algorithm to find the best quote. It works well here, but fails terribly there - in large part because an old, highly rated, but now outdated answer here is likely to be edited. bash.org doesn't allow editing, and it's not clear how you'd even go about editing decade-old jokes about now-dated internet memes even if you could... In any case, my point is that the right algorithm usually depends on the details of your problem. :-)
I know this question is quite old but I thought I'd contribute
I'd look at the TrueSkill system developed at Microsoft Research. It's like ELO but has a much faster convergence time (looks exponential compared to linear), so you get more out of each vote. It is, however, more complex mathematically.
http://en.wikipedia.org/wiki/TrueSkill
I don't like the Hot-or-Not style. Different people would pick different numbers even if they all liked the image exactly the same. Also I hate rating things out of 10, I never know which number to choose.
Pick A-or-B is much simpler and funner. You get to see two images, and comparisons are made between the images on the site.
These equations from Wikipedia makes it simpler/more effective to calculate Elo ratings, the algorithm for images A and B would be simple:
Get Ne, mA, mB and ratings RA,RB from your database.
Calculate KA ,KB, QA, QB by using the number of comparisons performed (Ne) and the number of times that image was compared (m) and current ratings :
Calculate EA and EB.
Score the winner's S : the winner as 1, loser as 0, and if you have a draw as 0.5,
Calculate the new ratings for both using:
Update the new ratings RA,RB and counts mA,mB in the database.
You may want to go with a combination.
First phase:
Hot-or-not style (although I would go with a 3 option vote: Sucks, Meh/OK. Cool!)
Once you've sorted the set into the 3 buckets, then I would select two images from the same bucket and go with the "Which is nicer"
You could then use an English Soccer system of promotion and demotion to move the top few "Sucks" into the Meh/OK region, in order to refine the edge cases.
Ranking 1-10 won't work, everyone has different levels. Someone who always gives 3-7 ratings would have his rankings eclipsed by people who always give 1 or 10.
a-or-b is more workable.
Wow, I'm late in the game.
I like the ELO system very much so, but like Owen says it seems to me that you'd be slow building up any significant results.
I believe humans have much greater capacity than just comparing two images, but you want to keep interactions to the bare minimum.
So how about you show n images (n being any number you can visibly display on a screen, this may be 10, 20, 30 depending on user's preference maybe) and get them to pick which they think is best in that lot. Now back to ELO. You need to modify you ratings system, but keep the same spirit. You have in fact compared one image to n-1 others. So you do your ELO rating n-1 times, but you should divide the change of rating by n-1 to match (so that results with different values of n are coherent with one another).
You're done. You've now got the best of all worlds. A simple rating system working with many images in one click.
If you prefer using the Pick A or B strategy I would recommend this paper: http://research.microsoft.com/en-us/um/people/horvitz/crowd_pairwise.pdf
Chen, X., Bennett, P. N., Collins-Thompson, K., & Horvitz, E. (2013,
February). Pairwise ranking aggregation in a crowdsourced setting. In
Proceedings of the sixth ACM international conference on Web search
and data mining (pp. 193-202). ACM.
The paper tells about the Crowd-BT model which extends the famous Bradley-Terry pairwise comparison model into crowdsource setting. It also gives an adaptive learning algorithm to enhance the time and space efficiency of the model. You can find a Matlab implementation of the algorithm on Github (but I'm not sure if it works).
The defunct web site whatsbetter.com used an Elo style method. You can read about the method in their FAQ on the Internet Archive.
Pick A-or-B its the simplest and less prone to bias, however at each human interaction it gives you substantially less information. I think because of the bias reduction, Pick is superior and in the limit it provides you with the same information.
A very simple scoring scheme is to have a count for each picture. When someone gives a positive comparison increment the count, when someone gives a negative comparison, decrement the count.
Sorting a 1-million integer list is very quick and will take less than a second on a modern computer.
That said, the problem is rather ill-posed - It will take you 50 days to show each image only once.
I bet though you are more interested in the most highly ranked images? So, you probably want to bias your image retrieval by predicted rank - so you are more likely to show images that have already achieved a few positive comparisons. This way you will more quickly just start showing 'interesting' images.
I like the quick-sort option but I'd make a few tweeks:
Keep the "comparison" results in a DB and then average them.
Get more than one comparison per view by giving the user 4-6 images and having them sort them.
Select what images to display by running qsort and recording and trimming anything that you don't have enough data on. Then when you have enough items recorded, spit out a page.
The other fun option would be to use the crowd to teach a neural-net.
I've read the book Programming Collective Intelligence and found it fascinating. I'd recently heard about a challenge amazon had posted to the world to come up with a better recommendation engine for their system.
The winner apparently produced the best algorithm by limiting the amount of information that was being fed to it.
As a first rule of thumb I guess... "More information is not necessarily better when it comes to fuzzy algorithms."
I know's it's subjective, but ultimately it's a measurable thing (clicks in response to recommendations).
Since most of us are dealing with the web these days and search can be considered a form of recommendation... I suspect I'm not the only one who'd appreciate other peoples ideas on this.
In a nutshell, "What is the best way to build a recommendation ?"
You don't want to use "overall popularity" unless you have no information about the user. Instead, you want to align this user with similar users and weight accordingly.
This is exactly what Bayesian Inference does. In English, it means adjusting the overall probability you'll like something (the average rating) with ratings from other people who generally vote your way as well.
Another piece of advice, but this time ad hoc: I find that there are people where if they like something I will almost assuredly not like it. I don't know if this effect is real or imagined, but it might be fun to build in a kind of "negative effect" instead of just clumping people by similarity.
Finally there's a company specializing in exactly this called SenseArray. The owner (Ian Clarke of freenet fame) is very approachable. You can use my name if you call him up.
There is an entire research area in computer science devoted to this subject. I'd suggest reading some articles.
Agree with #Ricardo. This question is too broad, like asking "What's the best way to optimize a system?"
One common feature to nearly all existing recommendation engines is that making the final recommendation boils down to multiplying some number of matrices and vectors. For example multiply a matrix containing proximity weights between users by a vector of item ratings.
(Of course you have to be ready for most of your vectors to be super sparse!)
My answer is surely too late for #Allain but for other users finding this question through search -- send me a PM and ask a more specific question and I will be sure to respond.
(I design recommendation engines professionally.)
#Lao Tzu, I agree with you.
According to me, recommendation engines are made up of:
Context Input fed from context aware systems (logging all your data)
Logical reasoning to filter the most obvious
Expert systems that improve your subjective data over the period of time based on context inputs, and
Probabilistic reasoning to do decision-making close-to-proximity based on weighted sum of previous actions(beliefs, desires, & intentions).
P.S.
I made such recommendation engine.