Window walk optimisation
You want to find a job that is near your home transport wise, near a supermarket, doctors that pays the most and has the cheapest house to buy or rent
We can walk different web sites in an iterative hill climbing algorithm.
Enumerate list of towns, cities and villages
Enumerate houses to rent and to buy at lowest cost sorted first.
Enumerate list of supermarkets near said houses and near job, sort by smallest distance
Enumerate jobs, sorted by most pay first
Now walk all these datasets as an optimisation problem to find a list of efficient combinations of jobs and housing.
This is far better than manually cross referencing data in separate tabs.
This is a dedicated site to this task
It could be a browser plugin and some GUI to build a walk search with the GUI.
In Excel we have a feature called What If which is very powerful. It can change cell values intelligently to find a goal value.
Well I thought of formalising a syntax for defining a hill walk that includes URLs.
If the servers do the work, then it requires the servers themselves change to support the protocol.
Metadrive could execute the data collection locally too. It would just result in more data transfer.
I see, but how will it walk? I think it would require something like the Metadrive to walk easily.
I think we had talked about the formats for describing target queries for conditions (for example, Ansible's roles define target states of machines, these roles are examples of objective functions -- a kind of queries for conditions to auto-satisfy), when we talked about the "wants files".
It's interesting, how program synthesis and program search boils down to walk optimization.
The goal is that any website tab can be cross referenced with another tab. So people can walk any information that is publicly available.
My previous post implied there would be a special way to offload walking to servers. I envision this being a common endpoint which lets you upload the walking problem as a set of variables and enumerations. You would have to provide the URL of other web pages too for the server to make requests on your behalf.
You would only be interested in the top N results - not the full result set as that would be lots of data to transfer.
You could combine this walk optimisation with the knapsack problem to schedule healthy meal recipes according to a calorie, fat, carbohydrate, protein budget. If you had a restaurant site with a menu or a recipe website or a grocery shopping website - they can all be combined!!
I envision using this feature to find hard drives that are simultaneously high performance, cheap and have low error rates. (Back blaze publishes hard drive failure stats)
I'd love this kind of thing. In fact, it doesn't end with the path -- take our dietary choices and nutrition -- they are a parameter in window walk optimization -- where to rent may depend on the kind of products and services we walk to, and that sequence of products and services to be engaged with in social circumstances may be sub-optimal -- why not to correlate them across different people (healthy or not), and suggest better ones (e.g., "Steve has been walking to gym, and his blood hemoglobin levels, and other parameters are good, and there are many such 'Steves' that I know, maybe you should consider that gym nearby your apartment?"), and then, based on them as a parameter, the "window walk optimization" would be adjusted.
In fact, walk-to product/service optimization would be easier if customers were to get the digital receipts (instead of paper ones) with data about their purchase details down to supply chain details about each product/service item purchased... It could be easily done preserving privacy (even for cash payments) with a URL (or QR code) printed on the paper receipt to fetch data details. It could easily be done with loyalty cards, that people use for discounts, not to speak with mobile banking apps (which, I guess, generally don't get the products details data). If it is not being done, the consumers are left data-poor. The category is: how to get back the shopping receipts data?
I thought about the practical concerns of executing the enumeration and walking of data.
I think the walking of data can be distributed for efficiency. It doesn't make sense to walk each record one at a time. It's a lot of data to download.
I think a continuation style can be used so that the provider of the data walks each sub iteration.
So the walking of data is continuations between collaborating servers.
Skihappy I suspect it can be used to find any multi faceted multi variable problem. If there is a database of communities then in theory those database providers should provide an API to interrogate the data.
The problem is Google doesn't categorise websites unless the website uses the term "community".
Good idea. However, I'd like to use this to find a community to join. Looking for a job is a very depressing proposition, but that's what the web is full of. It's very telling that no one is framing this issue in terms of community search. Granted, it'll do the same search using same or similar constraints, but using different lingo. Language is all we got to shape and express our thoughts