I am trying to perform ANN, but my data is split into partitions or "tenants." Searches are always restricted to a single tenant, which represents a small percentage of the total documents.
I first tried implementing this using a filter on a tenant string attribute. However, I encountered this piece of documentation, that suggests the performance will be poor:
There is a small problem here however. If the eligibility list is small in relation to the number of items in the graph, skipping occurs with a high probability. This means that the algorithm needs to consider an exponentially increasing number of candidates, slowing down the search significantly. To solve this, Vespa.ai switches over to a brute-force search when this occurs. The result is a efficient ANN search when combined with filters.
What's the best way to solve my problem? Will partitioning my data into separate namespaces trigger the creation of a separate HNSW graph per namespace?
Performance will be fine, the query planner will just choose to not use the ANN index for these queries. You'll find lots of details on this topic, including how to tune this, in this blog post: https://blog.vespa.ai/constrained-approximate-nearest-neighbor-search/
If all your queries are towards a single tenant which is a small percentage of the total documents I don't think you necessarily need to create an HNSW index at all, but this depends on the absolute numbers and the largest "small percentage".
(Namespaces are not relevant here - their only purpose is to safely add a string to ids so that you can have multiple sources of ids and still be guaranteed global uniqueness.)
Related
hope you are safe and well!
I have a question about regular or common ways of pair-matching if there is a database of users: say there are a few properties of each user, and when matching, each user could change the filtering options to only match those who fit their own requirement(so there is mutual selection between users), and we want to efficiently match 1000 users as precisely as possible.
For example, let's say there are 3 properties of every user: gender(female/male/other), study level(elementary/mediate/advanced), and grade(freshman/sophomore/junior/senior), and when matching, each user could choose to only match with people with their selected gender, study level and grade.
When focusing on 1 user, I could guess, on the perspective of database, we could use the filtering options in commands and get a list of those who satisfy both "my requirement" and "I fit their requirement"? However, I think this would be slow and asynchronous problems when there are 1000+ users in the matching phase at the same time?
I saw another post here discussed the blossom algorithm or greedy algorithm, which seem cool since if looking in a graph. Are they doable in this case? I guess if two users mutually fit both requirements, they would have an edge between the two nodes, and the value of edge could be comprehensive matching scores of 3 properties all together?
Anyway, I'm wondering is there a common way to do the pair matching precisely with at least 1000+ users at the same time?
Thank you so much!
If the requirement is that each match has to have the exact same properties, then the solution is fairly simple; just do a multiple criteria sort (ex. first sort by gender, then within each gender category sort by study level, etc.) and pair the identical users.
However, in a random dataset you're very unlikely to have perfect matches for all users. In that case you would want score pairs by how closely each category matches and use a more complex algorithm to maximize your overall matches. What you would do depends heavily on your use case and userbase size. Honestly, 1000 users is a very small number for modern computers; pretty much any polynomial time method (including blossom as you mentioned) would work fine.
The situation and the goal
Imagine a user search system that provides a proximity search from a user’s own position, which is specified by a decimal latitude/longitude combination. An Atlanta resident’s position, for instance, would be represented by 33.756944,-84.390278 and a perimeter search by this user should yield other users in his area from a radius of 10 mi, 50 mi, and so on.
A table-valued function calculates distances and provides users accordingly, ordered by ascending distance to the user that started the search. It’s always a live query, and it’s a tough and frequent one. Now, we want to built some sort of caching to reduce load.
On the way to solutions
So far, all users were grouped by the integer portion of their lat/long. The idea is to create cache files with all users from a grid square, so accessing the relevant cache file would be easy. If a grid square contains more users than a cache file should, the square is quartered or further divided into eight pieces and so on. To fully utilize a square and its cache file, multiple overlaying squares are contemplated. One deficiency of this approach is that gridding and quartering high density metropolitan areas and spacious countrysides into overlaying cache files may not be optimal.
Reading on, I stumbled upon topics like nearest neighbor searches, the Manhattan distance and tree-esque space partitioning techniques like a k-d tree, a quadtree or binary space partitioning. Also, SQL Server provides its own geographical datatypes and functions (though I’d guess the pure-mathematical FLOAT way has an adequate performance). And of course, the crux is making user-centric proximity searches cacheable.
Question!
I haven’t found much resources on this, but I’m sure I’m not the first one with this plan. Remember, it’s not about the search, but about caching.
Can I scrap my approach?
Are there ways of an advantageous partitioning of users into geographical divisions of equal size?
Is there a best practice for storing spatial user information for efficient proximity searches?
What do you think of the techniques mentioned above (quadtrees, etc.) and how would you pair them with caching?
Do you know an example of successfully caching user-specific proximity search?
Can I scrap my approach?
You can adapt your appoach, because as you already noted, a quadtree uses this technic. Or you use a geo spatial extension. That is available for MySql, too.
Are there ways of an advantageous partitioning of users into
geographical divisions of equal size
A simple fixed grid of equal size is fine when locations are equally distributed or if the area is very small. Geo locations are hardly equal distributed. Usually a geo spatial structure is used. see next answer:
Is there a best practice for storing spatial user information for
efficient proximity searches
Quadtree, k-dTree or R-Tree.
What do you think of the techniques mentioned above (quadtrees, etc.) and how would you pair them with caching?
There is some work from Hannan Samet, which describes Quadtrees and caching.
We have a database with hundreds of millions of records of log data. We're attempting to 'group' this log data as being likely to be of the same nature as other entries in the log database. For instance:
Record X may contain a log entry like:
Change Transaction ABC123 Assigned To Server US91
And Record Y may contain a log entry like:
Change Transaction XYZ789 Assigned To Server GB47
To us humans those two log entries are easily recognizable as being likely related in some way. Now, there may be 10 million rows between Record X and Record Y. And there may be thousands of other entries that are similar to X and Y, and some that are totally different but that have other records they are similar to.
What I'm trying to determine is the best way to group the similar items together and say that with XX% certainty Record X and Record Y are probably of the same nature. Or perhaps a better way of saying it would be that the system would look at Record Y and say based on your content you're most like Record X as apposed to all other records.
I've seen some mentions of Natural Language Processing and other ways to find similarity between strings (like just brute-forcing some Levenshtein calculations) - however for us we have these two additional challenges:
The content is machine generated - not human generated
As opposed to a search engine approach where we determine results for a given query - we're trying to classify a giant repository and group them by how alike they are to one another.
Thanks for your input!
Interesting problem. Obviously, there's a scale issue here because you don't really want to start comparing each record to every other record in the DB. I believe I'd look at growing a list of "known types" and scoring records against the types in that list to see if each record has a match in that list.
The "scoring" part will hopefully draw some good answers here -- your ability to score against known types is key to getting this to work well, and I have a feeling you're in a better position than we are to get that right. Some sort of soundex match, maybe? Or if you can figure out how to "discover" which parts of new records change, you could define your known types as regex expressions.
At that point, for each record, you can hopefully determine that you've got a match (with high confidence) or a match (with lower confidence) or very likely no match at all. In this last case, it's likely that you've found a new "type" that should be added to your "known types" list. If you keep track of the score for each record you matched, you could also go back for low-scoring matches and see if a better match showed up later in your processing.
I would suggest indexing your data using a text search engine like Lucene to split your log entries into terms. As your data is machine generated use also word bigrams and tigrams, even higher order n-grams. A bigram is just a sequence of consecutive words, in your example you would have the following bigrams:
Change_Transaction, Transaction_XYZ789, XYZ789_Assigned, Assigned_To, To_Server, Server_GB47
For each log prepare queries in a similar way, the search engine may give you the most similar results. You may need to tweek the similarity function a bit to obtain best results but I believe this is a good start.
Two main strategies come to my mind here:
the ad-hoc one. Use an information retrieval approach. Build an index for the log entries, eventually using a specialized tokenizer/parser, by feeding them into a regular text search engine. I've heard people do this with Xapian and Lucene. Then you can "search" for a new log record and the text search engine will (hopefully) return some related log entries to compare it with. Usually the "information retrieval" approach is however only interested in finding the 10 most similar results.
the clustering approach. You will usually need to turn the data into numerical vectors (that may however be sparse) e.g. as TF-IDF. Then you can apply a clustering algorithm to find groups of closely related lines (such as the example you gave above), and investigate their nature. You might need to tweak this a little, so it doesn't e.g. cluster on the server ID.
Both strategies have their ups and downs. The first one is quite fast, however it will always just return you some similar existing log lines, without much quantities on how common this line is. It's mostly useful for human inspection.
The second strategy is more computationally intensive, and depending on your parameters could fail completely (so maybe test it on a subset first), but could also give more useful results by actually building large groups of log entries that are very closely related.
It sounds like you could take the lucene approach mentioned above, then use that as a source for input vectors into the machine learning library Mahout (http://mahout.apache.org/). Once there you can train a classifier, or just use one of their clustering algorithms.
If your DBMS has it, take a look at SOUNDEX().
I am working on a possible architecture for an abuse detection mechanism on an account management system. What I want is to detect possible duplicate users based on certain correlating fields within a table. To make the problem simplistic, lets say I have a USER table with the following fields:
Name
Nationality
Current Address
Login
Interests
It is quite possible that one user has created multiple records within this table. There might be a certain pattern in which this user has created his/her accounts. What would it take to mine this table to flag records that may be possible duplicates. Another concern is scale. If we have lets say a million users, taking one user and matching it against the remaining users is unrealistic computationally. What if these records are distributed across various machines in various geographic locations?
What are some of the techniques, that I can use, to solve this problem? I have tried to pose this question in a technologically agnostic manner with the hopes that people can provide me with multiple perspectives.
Thanks
The answer really depends upon how you model your users and what constitutes a duplicate.
There could be a user that uses names from all harry potter characters. Good luck finding that pattern :)
If you are looking for records that are approximately similar try this simple approach:
Hash each word in the doc and pick the min shingle. Do this for k different hash functions. Concatenate these min hashes. What you have is a near duplicate.
To be clear, lets say a record has words w1....wn. Lets say your hash functions are h1...hk.
let m_i = min_j (h_i(w_j)
and the signature is S = m1.m2.m3....mk
The cool thing with this signature is that if two documents contain 90% same words then there is a good 90% chance that good chance that the signatures would be the same for the two documents. Hence, instead of looking for near duplicates, you look for exact duplicates in the signatures. If you want to increase the number of matches then you decrease the value of k, if you are getting too many false positives then you increase the number of k.
Of course there is the approach of implicit features of users such as thier IP addresses and cookie etc.
I'm looking for techniques to generate 'neighbours' (people with similar taste) for users on a site I am working on; something similar to the way last.fm works.
Currently, I have a compatibilty function for users which could come into play. It ranks users on having 1) rated similar items 2) rated the item similarly. The function weighs point 2 heigher and this would be the most important if I had to use only one of these factors when generating 'neighbours'.
One idea I had would be to just calculate the compatibilty of every combination of users and selecting the highest rated users to be the neighbours for the user. The downside of this is that as the number of users go up then this process couls take a very long time. For just a 1000 users, it needs 1000C2 (0.5 * 1000 * 999 = = 499 500) calls to the compatibility function which could be very heavy on the server also.
So I am looking for any advice, links to articles etc on how best to achieve a system like this.
In the book Programming Collective Intelligence
http://oreilly.com/catalog/9780596529321
Chapter 2 "Making Recommendations" does a really good job of outlining methods of recommending items to people based on similarities between users. You could use the similarity algorithms to find the 'neighbours' you are looking for. The chapter is available on google book search here:
http://books.google.com/books?id=fEsZ3Ey-Hq4C&printsec=frontcover
Be sure to look at Collaborative Filtering. Many recommendation systems use collaborative filtering to suggest items to users. They do it by finding 'neighbors' and then suggesting items your neighbors rated highly but you haven't rated. You could go as far as finding neighbors, and who knows, maybe you'll want recommendations in the future.
GroupLens is a research lab at the University of Minnesota that studies collaborative filtering techniques. They have a ton of published research as well as a few sample datasets.
The Netflix Prize is a competition to determine who can most effectively solve this sort of problem. Follow the links off their LeaderBoard. A few of the competitors share their solutions.
As far as a computationally inexpensive solution, you could try this:
Create categories for your items. If we're talking about music, they might be classical, rock, jazz, hip-hop... or go further: Grindcore, Math Rock, Riot Grrrl...
Now, every time a user rates an item, roll up their ratings at the category level. So you know 'User A' likes Honky Tonk and Acid House because they give those items high ratings frequently. Frequency and strength is probably important for your category aggregate score.
When it's time to find neighbors, instead of cruising through all ratings, just look for similar scores in the categories.
This method wouldn't be as accurate but it's fast.
Cheers.
What you need is a clustering algorithm, which would automatically group similar users together. The first difficulty that you are facing is that most clustering algorithms expect the items they cluster to be represented as points in a Euclidean space. In your case, you don't have the coordinates of the points. Instead, you can compute the value of the "similarity" function between pairs of them.
One good possibility here is to use spectral clustering, which needs precisely what you have: a similarity matrix. The downside is that you still need to compute your compatibility function for every pair of points, i. e. the algorithm is O(n^2).
If you absolutely need an algorithm faster than O(n^2), then you can try an approach called dissimilarity spaces. The idea is very simple. You invert your compatibility function (e. g. by taking its reciprocal) to turn it into a measure of dissimilarity or distance. Then you compare every item (user, in your case) to a set of prototype items, and treat the resulting distances as coordinates in a space. For instance, if you have 100 prototypes, then each user would be represented by a vector of 100 elements, i. e. by a point in 100-dimensional space. Then you can use any standard clustering algorithm, such as K-means.
The question now is how do you choose the prototypes, and how many do you need. Various heuristics have been tried, however, here is a dissertation which argues that choosing prototypes randomly may be sufficient. It shows experiments in which using 100 or 200 randomly selected prototypes produced good results. In your case if you have 1000 users, and you choose 200 of them to be prototypes, then you would need to evaluate your compatibility function 200,000 times, which is an improvement of a factor of 2.5 over comparing every pair. The real advantage, though, is that for 1,000,000 users 200 prototypes would still be sufficient, and you would need to make 200,000,000 comparisons, rather than 500,000,000,000 an improvement of a factor of 2500. What you get is O(n) algorithm, which is better than O(n^2), despite a potentially large constant factor.
The problem seems like to be 'classification problems'. Yes there are so many solutions and approaches.
To start exploration check this:
http://en.wikipedia.org/wiki/Statistical_classification
Have you heard of kohonen networks?
Its a self organing learning algorithm that clusters similar variables into similar slots. Although most sites like the one I link you to displays the net as bidimensional there is little involved in extending the algorithm into a multiple dimension hypercube.
With such a data structure finding and storing neighbours with similar tastes is trivial as similar users should be stores into similar locations (almost like a reverse hash code).
This reduces your problem into one of finding the variables that will define similarity and establishing distances between possible enumerate values ,like for example classical and acoustic are close toghether while death metal and reggae are quite distant (at least in my oppinion)
By the way in order to find good dividing variables the best algorithm is a decision tree. The nodes closer to the root will be the most important variables to establish 'closeness'.
It looks like you need to read about clustering algorithms. The general idea is that instead of comparing every point with every other point each time you divide them in clusters of similar points. Then the neighborhood may be all the points in the same cluster. The number/size of the clusters is usually a parameter of the clustering algorithm.
Yo can find a video about clustering in Google's series about cluster computing and mapreduce.
Concerns over performance can be greatly mitigated if you consider this as a build/batch problem rather than a realtime query.
The graph can be statically computed then latently updated e.g. hourly, daily etc. to then generate edges and storage optimized for runtime query e.g. top 10 similar users for each user.
+1 for Programming Collective Intelligence too - it is very informative - wish it wasn't (or I was!) as Python-oriented, but still good.