Seriate

To seriate your data, choose Seriate | Seriate from the OptiPath main menu. This will bring up the OptiPath Seriation dialog. Some summary information about your data is presented. This includes the numbers of artifacts, assemblages and features in your data. You are also shown the earliest and latest allowed dates for this seriation.

Clicking the Compute button will start the seriation.

Seriate Menu

The Seriate menu contains a number of items:

Seriate -  initiates the calculation of a new seriation. There is a corresponding Seriate button on the toolbar.

Reverse Order - reverses the order of the current seriation. Seriation is an ordering technique, not a dating technique. Optimal Path seriation is able to assign relative dates to the items being seriated; however, this depends upon the dates you enter as the Earliest and Latest dates for the seriation. Even then, without Earliest and Latest dates for each artifact (assemblage), OptiPath is not able to orient the direction of the relative dating - i.e. it does not know which end is earliest and which is latest. The Reverse command allows you to reverse the order of the current seriation, as seen in the Results table. There is a corresponding Reverse button on the toolbar.

Ignore User Dates - tells OptiPath to ignore the user entered Earliest and Latest dates for each artifact (assemblage). This option has two purposes. First, it allows you to see how OptiPath would seriate your items without having to go through all the items and delete their Earliest and Latest dates (and re-enter them later when you want to use them again). Second, if you know the dates of your items, you can use the Results | Errors menu item (which is available only if you have selected Seriate | Ignore User Dates) to prepare an analysis for you.

Seriate Dialog

There are a number of fields associated with the seriation dialog:

Features - the number of features being considered in the seriation. This is for your information only, the number of features can be modified in the Features table.

Artifacts - the number of artifacts being considered in the seriation. This is for your information only, the number of artifacts can be modified in the Artifacts table.

Assemblages - the number of assemblages being considered in the seriation. This is for your information only, the number of assemblages can be modified in the Seriations table.

Earliest Date - the earliest date that will be assigned to an artifact in the seriation. This is for your information only, the earliest date can be modified in the Seriations table.

Latest Date - the latest date that will be assigned to an artifact in the seriation. This is for your information only, the latest date can be modified in the Seriations table.

Start - the length of the path (the value of the objective function to be minimized) at the start of the seriation. This is for your information only.

Current - the length of the path (the value of the objective function to be minimized) for the current seriation. This is for your information only.

Improvement - the improvement in the path length (the value of the objective function to be minimized) between the current and the starting seriation. This is for your information only.

There is a single editable parameter on the Seriation dialog:

Effort - the amount of effort you want the algorithms to devote in finding an optimal answer (minimum path length). The seriation problem is modeled as a special case of the "traveling salesman problem", which is notoriously difficult to solve. A seriation of a few dozen artifacts could take an inordinate amount of time (hours or days) to solve optimally on a desktop computer. OptiPath uses an heuristic optimization algorithm that can run quickly although it is not guaranteed to find the very best possible answer. For large samples, even a heuristic can require substantial amounts of computation time. The Effort parameter allows you to control the amount of computation time: the smaller the value of Effort, the quicker the computation and possibly the less optimal the answer. A computation can be suspended or abandoned in mid-stream without losing the benefit of the work done so far. Running the computations repeatedly (perhaps with different levels of Effort) can lead to improved results. If you have specified Earliest or Latest times for individual artifacts on the Artifacts table, it will probably take much longer for the algorithms to find a good answer and you may want to use a smaller Effort value.

There is a single check box on the Seriation dialog:

Randomize - the heuristic solution technique will normally get the same answer each time you run it, provided you start with the same seriation each time. However, if you check the Randomize box, the heuristic will modify its search somewhat and may find a different answer. Since the heuristic finds a good solution, not necessarily the best, this provides a means for extending a search for a best answer, by doing multiple seriations with Randomize checked. Randomize is a random process and you may not be able to repeat the results you get using Randomize. If you want OptiPath to try alternative paths that can be replicated, use the Seed parameter on the Seriations dialog. The difference between Randomize and Shuffle (below) is that Shuffle may result in a worse solution (since it starts the search process all over again from a randomly generated starting point) whereas Randomize will not make things worse (since it starts from your current seriation, only randomizing the process of searching for an improvement).

Finally, there are three buttons on the Seriation dialog:

Shuffle - the heuristic solution technique will normally get the same answer each time you run it, provided you start with the same seriation each time and have not checked the Randomize box. However, if you start from a different seriation, you may get different results (the heuristic gets a good solution, not necessarily the best). Consequently, you may be able to find a better answer by starting from a different solution. Shuffle allows you to do this. Shuffling will replace the current seriation with a random one (which won't be very good in itself but may allow the heuristic to find a better solution). When you click the Shuffle button a warning dialog will appear to remind you that shuffling will replace your current seriation with a randomly generated one. Click OK to continue. Shuffle, like Randomize, is a random process and you may not be able to repeat the results you get using Shuffle. The difference between Randomize (above) and Shuffle is that Shuffle may result in a worse solution (since it starts the search process all over again from a randomly generated starting point) whereas Randomize will not make things worse (since it starts from your current seriation, only randomizing the process of searching for an improvement).

Compute - the Compute button initiates a new seriation, and then turns into a Pause button. During the computation there will be a progress bar. The heuristic algorithms are iterative. If they make a significant improvement on the solution, they may try again and again until they stop making improvements. You can pause the algorithms with the Pause button or stop it altogether with the Stop button without losing the benefit of the work done so far. At any time the dialog shows the value of the starting solution (4.91 here) and the best solution found so far (4.47) and expresses the improvement as a percentage (9.01%). If you have specified Earliest or Latest times for individual artifacts on the Artifacts window, it will probably take much longer for the algorithms to find a good answer.

When the algorithm is finished you will be given the computation time. Without going back to the OptiPath window you can run the algorithms again if you want to see if they can make any further improvements.

Exit - the Exit button will take you back to the main OptiPath window where you can view the results.