4.1 Database

[14]We used Neo4j Desktop, version 1.6.0, to store and analyse the data. The database management system (DBMS) had version 5.20.0. In our case, the use of a graph database offers several advantages over relational databases. Since the following data analyses are network analyses, it makes sense to store them as graphs. Graph algorithms make it easy to explore the relationships in the graph.‍[41] Because graph databases focus much more on relationships, they can be queried using the Cypher programming language without the need for complicated SQL joins that are required in relational databases. Neo4j also offers good options for data visualisation as graphs with ›Neo4j Bloom‹. The use of graph algorithms in data analysis is also made possible in Neo4j. We used the library ›Graph Data Science Library‹, Version 2.7.0.

4.2 Data Structure and Cleaning

[15]Both museums provide a large amount of metadata about the objects and artists.‍[42] Important information included in our study encompasses the categories we named »ObjectID« and »Accession Year«, »Creditline« and »Department«. Entries without a creditline were excluded from the study. Hence, all results mirror the present level of data accuracy within the collections. It is essential for understanding the results to point out that the classification of the objects into departments reflects their current structure. The initial assignment of objects immediately after receipt cannot be traced using the data provided. This would require intensive research in the respective physical archives. Accordingly, more recent departments such as the »Young V&A Collection« are also reflected in our evaluations. In the next step, the data was filtered and sorted according to the period under investigation (1880 to 1910) and the object of investigation (donations). These measures were carried out with a spreadsheet application in Microsoft Excel, as it is easy to filter for entries in table columns. The columns »Creditline«, »Department«, »ObjectID« and »Accession Year« were used. For example, the »Creditline« column was searched for the keywords »Bought«, »Purchase«, »Commissioned« and »Acquired«. These entries were then deleted. First, we checked that only creditlines that clearly denote gifts (keywords: »Bequest«, »Bequeathed from«, »Given by«, »Gift of«, »Received from«) were retained. These words were then removed in order to standardise the data, leaving only the names.

[16]Further data cleansing proved to be difficult. As Neo4j cannot merge data into one node if the spelling differs even slightly (e.g. by spaces), the entries had to be standardised. The names of the patrons were often spelt differently or were abbreviated. For example, the art collector George Salting was entered as »George Salting, Esq.«, »Salting«, »G. Salting«, »Mr George Salting«, »Mr George Salting, Esq.«, »George Salting.« and our preferred spelling »George Salting«. Other persons such as Joseph Henry Fitzhenry are noted in different variations of their first name initials: »J H Fitzhenry«, »J.H. Fitzhenry«, »J, [sic] H Fitzhenry«, »J Fitzhenry«. First names are sometimes abbreviated as initials, other times written out in full: »D. B. Myers« and »Dudley B. Myers«. Although women are usually referred to by their own first names, there are also cases in which they are addressed by their husband’s first and last name. Here, only the addition of »Mrs«. indicates that it is the wife: »Mr. and Mrs. Robert de Forest« instead of »Robert and Emily de Forest«. While the addition of »Mr.« could easily be removed, female forms of address had to be retained. All other additions such as military titles, year, place of residence, memorial information (in memory of...) and the honorary title »Esq.« were removed for data unification. Punctuation marks like commas and full stops were also deleted. To avoid differences due to capitalisation, the data was transformed to lower case. The different spelling of names and spelling mistakes were another problem we had to tackle: »Margarette A. Jones« and »Margaretta A. Jones«. Using fuzzy matching in python, different spellings were identified. We used the package FuzzyWuzzy‍[43] with a ratio of > 85. The new list was then reexamined by the authors and the names adjusted. Although isolated errors cannot be ruled out, this data was sufficiently streamlined to proceed with network analysis.

4.3 Network

[17]The basic structure of the network is identical for both museums. We implemented four different types of nodes: ›Creditline‹ (the donor), ›Department‹ (the museum department to which the object is currently assigned), ›ObjectID‹ (the ID of the individual object) and ›AccessionYear‹ (the year in which the object was inventoried). Each node has a property (›credname‹, ›depname‹, ›ID‹ and ›year‹) in which the name or number is stored. The nodes are connected through various directed, simple edges: (Object → Department), (Object → Creditline), (Creditline → AccessionYear), (Creditline → Department), (Object → AccessionYear). Unsurprisingly, given the outlined data structure, our graphs fulfil the conditions of a small world network.‍[44] This type of network is characterised by a high clustering coefficient and a low average shortest path length.‍[45] The graphs of both museums fulfil these conditions, allowing us to conclude that they form a highly interconnected network with short paths.

[18]In order to use graph algorithms in Neo4j, the graph was first converted into a bipartite network‍[46] with the node types ›Department‹ and ›Creditline‹ in a new DBMS instance. To ensure that the information on the number of objects assigned to a creditline is not lost, this number was added to the respective creditlines as node weight. The newly created network is an affiliation network. Sociologist Katherine Faust defines this as follows: »networks of actors tied to each other through their participation in collectivities, and collectivities linked through multiple memberships of actors. [...] Formally, an affiliation network consists of two key elements: a set of actors and a collection of subsets of actors (called events)«.‍[47] In our case, the creditlines are the actors, some of which are assigned to several events, i.e. departments. In order to perform the algorithm calculations, the graph in Neo4j must be projected onto a unimodal network.‍[48] To do this, an edge was created between two creditlines whenever both donors donated to the same department and were thus connected. If the two donors have donated to several identical departments, the edge is given a weight that is increased by one per department. The department nodes were then removed from the network.

7.1 Objects per Department

[26]Based on our data, which consists only of the objects available online, a total of 21,041 objects were donated to ten departments of the V&A during the period under review. These objects form our data basis. The most objects by far were given to the department »Prints, Drawings and Paintings«, followed by the »Ceramics Collection«. Together, these two departments dominate the donations to the museums with around 65 % of objects. With 2,846 objects, the »East Asia Collection« is in third place. The two departments »East Asia Collection« and »South & South East Asia Collection« were viewed together to make them more comparable with the much more general »Asian Art« department in the Met. The joint department has 3,811 objects but remains in third place.

Fig. 1: Amount of objects per V&A department. [Chart: Astrid Brixy 2024]

[27]Based on the equivalent data source at the Met (i.e., the data available online), 11,566 objects were donated to the department »European Sculpture and Decorative Arts«. The two departments »Asian Art« (6,443 objects) and »Egyptian Art« (6,028 objects) follow at considerable distance. The smallest departments were the then still young field of photography (2 objects) and the department showcasing medieval European art »The Cloisters« (17 objects).

Fig. 2: Amount of objects per Met department. [Chart: Astrid Brixy 2024]

7.2 Creditlines per Department

[28]The »Prints, Drawings and Paintings« and »Ceramics Collection« departments at the V&A also dominate in terms of the number of creditlines per department (267 and 151 creditlines). Other departments have far fewer donors: only 48 creditlines are attested to the »Middle East Section«, which is fourth place in terms of the number of objects.

Fig. 3: Amount of creditlines per V&A department. [Chart: Astrid Brixy 2024]

[29]As with the V&A, the department with the most objects at the Met corresponds to the one with the largest number of patrons (»European Sculpture and Decorative Art«, 244 creditlines). »The American Wing«, i.e. American art, especially paintings, is in second place by only a small margin (220 creditlines). As somewhat expected, the smallest departments in terms of object donations are also those with the smallest number of donors.

Fig. 4: Amount of creditlines per Met department. [Chart: Astrid Brixy 2024]

7.3 Development of Object Donations over Time

[30]The number of objects donated to the V&A varies considerably over the period analysed. The years 1909 (2,041 objects), 1885 (2,557 objects) and above all 1910 (4,136 objects) stand out with a particularly high number of donations. The fewest objects were donated in 1884 (37 objects), 1892 (25 objects), 1895 (23 objects) and 1883 (11 objects). A comparison of the number of objects donated by decade reveals the dominance of the 1900s: 12,277 objects, i.e. approximately 58 % of the total number of objects were donated then. 25 % (5,329 objects) were donated in the 1880s, while in the 1890s only 3,435, i.e. approximately 16 % of the objects, were donated to the museum.

Fig. 5: Amount of objects per year in the V&A. [Chart: Astrid Brixy 2024]

[31]The years 1908 to 1910 are the most popular years in the Met, followed by 1889 and 1907. With only nine objects, hardly any donations were made to the museum in 1884. Overall, approximately 59 % of the objects are from the 1900s, while the 1880s (approximately 19.9 %) and 1890s (approximately 21 %) resemble each other.

Fig. 6: Amount of objects per year in the Met. [Chart: Astrid Brixy 2024]

7.4 Major and Minor Donors

[32]The structure of major and minor donors in both museums shows striking similarities. We consider people who donated up to ten objects as minor donors and all larger donations as major donors. In total, 630 people donated 21,041 objects to the V&A and 821 people donated 39,450 objects to the Met.

[33]In the V&A, a clear trend can be identified in the distribution of the number of objects per creditline: most donors form a cluster with only few object donations. There are more major donors with donations up to over 3,000 objects in the V&A than in the Met. Interestingly, no donors have given between 1,500 and 2,000 objects to the museum.

Fig. 7: Amount of objects per amount of creditlines in the V&A. [Chart: Astrid Brixy 2024]

[34]At the Met, too, most people donated only a few objects. The larger the number of objects, the fewer and more isolated are the donors. One conspicuous outlier with 10,168 object donations is the Rogers Fund. This monetary fund was donated to the museum in 1903 by railway manufacturer Jacob S. Rogers. In the case of purchases with money from the fund, this was always indicated as a creditline and classified as a donation.

Fig. 8: Amount of objects per amount of creditlines in the Met. [Chart: Astrid Brixy 2024]

[35]A closer look at the results reveals that almost 50 % of donors in both museums gave only one object each to the museum. Almost all patrons donated a maximum of 1,000 objects to the museum, with very few larger donations.

[36]An analysis of the relationship between the number of departments and the size of donations shows that even among major donors, there is a clear trend of donating to just one department. However, the number of people donating to multiple departments is also increasing.

7.5 Degree Centrality and Roles of the Creditlines

[37]To find out which donors are particularly central in the network, i.e. donated a lot and to different departments, is important, as these donors substantially shaped the collections through their donations. Therefore, the degree centrality of the creditlines was calculated, then the number of objects was added. The 50 creditlines with the highest scores (51 in the Met, as there are two creditlines tied for 50th rank) were identified. Then, the associated persons were analysed biographically. The donors were categorised and assigned to at least one of the following classes (the classes were created for this purpose by the authors): middle class, upper class, artist or craftsman, museum employee / professional, military member, church functionary, aristocrat, person only donating the estate of a relative (not a collector themselves), person unknown (no information found) and institution / association / company. Some of the people were also assigned several roles.

[38]The evaluation shows that the hubs in the Met are severely dominated by members of the upper class. In the V&A, upper class members also dominate the hubs. However, it is difficult to make an exact statement, as 14 people, and thus the largest group, could not be identified. In contrast, more people could be identified in the Met, only 7 remain unknown.

7.6 Betweenness Centrality of the Creditlines

[39]The 50 patrons with the highest betweenness centrality‍[52] include many already identified hubs in both museums. The bridges were identified in two complementary ways. First, betweenness centrality was calculated on the unimodal creditline projection. Second, using the original bipartite creditline–department data, creditlines that donated to exactly two departments, each shared by no more than four other creditlines, were identified. This step ensures that bridges not only have a central network position but also connect departments that are otherwise weakly linked. It is noticeable that all the department combinations each contain at least one department that was given few objects. The most noteworthy patrons in the V&A with high betweenness centrality, who also connect otherwise almost disjoint departments, are first and foremost Joshua Dixon, but also George Salting, John Jones, Mrs Amelia Vertue Jodrell, Edmond Dresden, Colonel F. R. Waldo-Sibthorp, Robert Taylor, George Donaldson and Thomas Wardle. In the Met, there is a similar accumulation of creditlines with high degree and betweenness centrality. However, there are considerably more almost disjoint department combinations (156), so that no individual persons will be named.

7.7 Louvain Modularity Algorithm of the Creditlines

[40]The Louvain algorithm was used to find communities of creditlines. The creditlines can be assigned to only four communities in the Met, but six communities in the V&A. These communities are defined by the departments to which the creditlines have donated, with each community being dominated by specific departments.

[41]In the Met, cluster 1 consists of donations to a maximum of four departments, always to »The American Wing«. Cluster 2 includes donations to at most six departments, including »Drawings and Prints«, »European Paintings« or »Musical Instruments«. Donations to up to five departments, including »European Sculpture and Decorative Arts« or the »Costume Institute« are collected in Cluster 3. All other departments and all creditlines that have donated to more than six departments form Cluster 4.

[42]In the V&A, Cluster 1 contains the »Metalwork Collection« (up to 2 departments), Cluster 2 the »South & Southeast Asia Collection« (up to 6 departments) and Cluster 3 the department »Prints, Drawings & Paintings« (up to 3 departments). Cluster 4 either includes the »Sculpture Collection«, the »Textiles and Fashion Collection« or the »East Asia Collection« (up to 9 departments), while Cluster 5 contains the »Middle East Section« and »Ceramics Collection« (up to 7 departments). Cluster 6 includes the departments »Furniture and Woodwork Collection« and »Young V&A Collection« (up to 4 departments).

7.8 Degree Centrality of the Departments

[43]The degree centrality of the departments, i.e. the number of donors who have donated to this department, was calculated using the algorithms of the Neo4j Graph Data Science Library on a bipartite network without adding the node weights, i.e. the number of objects. Otherwise, the result would not differ from the number of objects per department. The department with the highest degree is »Ceramics Collection«, followed by »Metalwork Collection«. The »Young V&A Collection« has a degree of only 70.

[44]The Met is dominated by the department with the most object donations, »European Sculpture and Decorative Art«. The departments »The American Wing« and »Islamic Art« follow at some distance. With »Photographs«, the department with the fewest objects is also the one with the smallest degree.

8.1 Objects per Department

[46]The department with the most object donations in the V&A is the »Prints, Drawings and Paintings Collection«. It contains large donations and bequests from a number of artists. In 1887, Isabel Constable, daughter of the painter John Constable, gave the museum »95 paintings, 297 drawings and three sketchbooks«‍[53] of her late father. However, no further information can be found about the artist Louis van Pelégheus, from whom the V&A received 1,129 works. Some collectors such as the Reverend Chauncy Hare Townsend are also among the major donors, although Townsend died in 1869, so a later year was incorrectly given as the »Accession Year«. With the founding of the Tate Gallery in 1908, the V&A no longer collected paintings and drawings.‍[54]

[47]The Met obtained a large number of donations to the »European Sculpture and Decorative Arts« department. The department can be seen as a mixture of the »Sculpture Collection« and »Furniture and Woodwork« at the V&A. This can be explained by the importance of European art for the status of New York’s high society as a distinction from other social classes. Enthusiasm for European culture was rooted in the European travels of the wealthy elite. The significance of art collections and the strong influence of Europe are also reflected in the architecture and interior design of the homes of the high society. Many European furnishings, including furniture, became part of the Met collection as ›decorative art‹ and thus found their place in the »European Sculpture and Decorative Arts« department.‍[55]

[48]At the V&A, the second most popular department is the »Ceramics Collection«. There are two main reasons for this: Lady Charlotte Schreiber and archaeological excavations in Egypt. Lady Charlotte Schreiber was a highly educated woman. She shared a passion for ceramics with her second husband, Charles Schreiber. They lent the V&A parts of their collection several times and travelled through Europe 24 times in search of new objects.‍[56] After the death of her husband in 1884, she donated 1,800 objects to the V&A in 1885. Thus, the museum came into possession of the »world’s largest collection of 18th century English porcelain, earthenware, glass and enamels«‍[57]. It is worth mentioning that this was only a small part of her ceramics collection, totalling around 12,000 objects. Another reason for the department’s appeal to donors was excavations in Egypt by the Egypt Exploration Fund (EEF). Founded in 1882, the organisation donated many objects to the V&A. In the absence of a separate department, most of the objects were placed in the »Ceramics Collection«, but some also in »Metalwork« and the »Middle East Section«.‍[58] The Met also received large donations from the EEF, which explains the size of their department of »Egyptian Art«.

[49]The Met’s department of »Asian Art« was originally founded in 1915 as »department of Far Eastern Art«. Today, it houses one of the largest and most comprehensive collections of Asian art in the world, with over 35,000 objects.‍[59] Even before the department was founded, the Met already owned many Asian artworks. The V&A also obtained a huge amount of Asian artefacts in our survey period. Most Asian artworks are of Japanese or Chinese provenance, between which no distinction was made for a long time. The global political situation explains the phenomenon of an increasing acquisition of Asian objects. From the beginning of the Opium War in 1839, Chinese and Japanese art became a focus of British and American public attention. After the end of the war in 1841, trade between the countries became much easier.‍[60] From the 1860s onwards, supposedly Japanese objects were particularly popular, but above all Chinese blue and white porcelain plates. However, it was not until the 1876 World’s Fair in Philadelphia that Asian art achieved a breakthrough in New York high society’s passion for collecting.‍[61] The most important single donor of Asian art to the V&A during the period under study was George Salting. His bequest in 1909 included over 200 of the plates mentioned above.‍[62] For the Met, the patrons Stephen Whitney Phoenix, Samuel Colman, Charles Stewart Smith and Edward C. Moore should be mentioned.‍[63] It is striking that there are nine departments of the V&A to which not a single object was donated. The museum’s collection focus expanded considerably after the end of the period under review. For some departments, this can be clarified by taking a look at the founding history of the departments. For example, the »Wedgwood Collection«, the collection of the potter and entrepreneur Josiah Wedgwood (1730–1795), was only donated to the museum in 2014. That no patrons contributed photographs to the museum can be attributed to the fact that that technology was still in its infancy at the time. This is also consistent with the result from the Met. Here, the Photography Department also only received 2 objects.

8.2 Development of Object Donations over Time

[50]Surprisingly, the correlation between the number of donations and the respective years in the Met does not show any clear causality between customs duties and object donations. In the years 1883 to 1891, 30 % customs duties had to be paid on the import of European objects to the US in order to promote its own art scene. Despite this, there was no permanent drop in object donations. Even the duty-free years between 1895 and 1897 and the reintroduction of 15–20 % duty in 1897 are not reflected in the donations. Only the large increase in donations in 1910 could be linked to the cancellation of customs duties on works of art older than 20 years in 1909.‍[64] One explanation could also be our data foundation, with a possible lack of donated objects in our time frame which might falsify our results. However, the probability that the data from 1883 to 1891 is significantly more complete than that of other years is low. Accordingly, a different reason must be assumed, and further research is needed.

[51]In the case of the V&A, the year 1910 is particularly noticeable with an accumulation of central patrons. These patrons also stand out due to their high degree centrality which means that they also donated particularly extensively. In 1910, Joseph Henry Fitzhenry donated his »collections of French porcelain and Dutch faience«‍[65] to the museum. Sydney Vacher, Robert Forrer and Alan Cole are also listed. The museum came into possession of two large estates as well. George Salting died in December 1909 and left parts of his impressively large collection to the museum. Captain Henry Boyles Murray also bequeathed »his collection including porcelain and metalwork to the Museum together with £50,000 for future acquisitions«‍[66] in 1910. In 1885, the large donation of ceramics from Lady Charlotte Schreiber was given to the museum. The Egypt Exploration fund donated many objects in 1909. Overall, an especially large number of different patrons donated objects that year.

8.3 Major and Minor Donors

[52]The large number of minor donors in both museums indicates a similar distribution of patrons from the upper middle class compared to the wealthy members of the upper class. The V&A was founded with a much higher educational goal for the population and, as an arts and crafts museum, focuses on craftsmanship, making the similarities in distribution even more striking. The initial hypothesis derived from the literature that the V&A attracted more minor donors than the Met was not confirmed.

8.4 Degree Centrality and Roles of the Creditlines

[53]Examining the class affiliation of patrons with the highest degree centrality of both museums confirms the hypothesised differences. The assumption that there are more aristocrats among the patrons of the V&A is therefore correct. In addition, the V&A has more bequests that were given to the museum by surviving relatives. This also confirms our hypothesis that the social status of donors at the V&A is much more diverse than at the Met, which is clearly dominated by members of high society (51 % at the Met vs. 22 % at the V&A). We assume a connection with the V&A’s focus as an arts and crafts museum. Thus, artisans who wanted to demonstrate their own skills donated individual objects to the museum. In addition, the New York high society purposefully excluded the middle and lower class. Accordingly, they did not donate any objects. However, it must be pointed out that 24 % of the hubs could not be biographically identified in the V&A and therefore could not be effectively included in the analysis. In the Met, this proportion is 10 % lower. The analysis of the social background of the hubs confirms our assumption that the Met was much more strongly dominated by high society than the V&A. This finding is also in line with the historical backgrounds of both museums, as explained at the beginning.

8.5 Betweenness Centrality of the Creditlines

[54]Among the donors with a high betweenness centrality, many hubs can be found. Major donors often had more than one specific field of interest and therefore gave objects to several departments. The fact that significantly more departments are connected by bridges in the Met can be seen as an indication of the less focused interests of patrons there compared to British donors.

8.6 Louvain Modularity Algorithm of the Creditlines

[55]Overall, the V&A has more differentiated department clusters. Interestingly, two clusters show great similarities in both museums. »Prints, Drawings and Paintings« is a cluster at the V&A, and »Drawings and Prints« and »European Paintings« are also a cluster at the Met. Only paintings from »The American Wing« form a separate group. One can speculate that donations to »The American Wing« were also made from a patriotic-nationalist background and that some of these donors may have wanted to criticise the popularity of European art in the US with their donations. Accordingly, this could be a community with different behaviour than the donors of the other departments. It can therefore be assumed that US patrons with an interest in European art and British patrons with the same interest have a more similar pattern than donors who have given American art to the Met. The same applies to patrons of the departments »European Sculpture and Decorative Arts« and »The Costume Institute« at the Met and the departments »Sculpture Collection« and »Textiles and Fashion Collection« at the V&A, which also belong to the same community, respectively.

8.7 Degree Centrality of the Departments

[56]The department with the most objects in the V&A, »Prints, Drawings and Paintings«, only has the fourth highest degree centrality score. This shows that many creditlines donated exclusively to this department. Thus, it has fewer connections to the other departments. At the same time, a particularly large number of people donated to this department, which indicates a large number of people interested exclusively in classical art. In contrast, the department with the second-largest number of objects, the »Ceramics Collection«, is characterised by a higher number of creditlines who were interested in more diverse areas and who also donated to other departments. Interestingly, there is a particularly close connection with the »Metalwork Collection«. Both departments are especially suitable for small donations, as small objects such as cups were easily available. In the Met, on the other hand, patrons of »Asian Art« have a one-sided special interest, while donors of »European Sculpture and Decorative Arts« and »The American Wing« usually also donated to other departments. At the same time, both departments have a particularly high number of creditlines, so it can be assumed that most patrons of the Met also donated to these departments.

8.8 Graph Distances

[57]Comparison between entire graphs is still at a somewhat experimental stage.‍[67] Nevertheless, the distance between two graphs can be used as an indicator of similarity; different distance measures compute different feature distances. As seen above, the general donor structure of the Metropolitan Museum and the Victoria and Albert Museum seems to be rather similar. This should be reflected in similar results for network distances. However, the networks are also quite different in size which can skew outcomes. We used the python library netrd‍[68] to compute the distances described below. For better comparison, we also created two random Erdös-Rényi graphs with NetworkX’s fast_gnp_random_graph() function.‍[69] These were set up to have the same number of nodes and a similar number of edges to the respective real world museum networks and are given as »Random (V&A)« and »Random (Met)«. The probability for edge creation was determined experimentally to match the number of edges in the museum graphs. All results are rounded to three decimal places. The algorithms used were Laplacian spectral distance and NetSimile as implemented in the netrd library. The Laplacian spectral distance takes the Laplacian matrix, consisting of the graph’s adjacency matrix and its diagonal matrix of vertex degrees, and computes the matrix spectra or its set of eigenvalues; in the following steps the spectra are converted to be continuous and the resulting distributions can be compared to compute the distance between graphs.‍[70] NetSimile takes a larger set of features per node (degree, clustering coefficient, two-hop away neighbours, average clustering coefficient of neighbours, edges in the node’s ego-net, outgoing edges from the node’s ego-net, number of neighbours of the node’s ego-net) and aggregates over the resulting feature-node matrix.‍[71] The vectors computed in the last step can subsequently also be compared to calculate graph distance. The authors of NetSimile note that one of its advantages is that feature extraction seems to be somewhat independent of network size.‍[72]

[58]Both distance measures indicate a structural similarity between the two museum graphs as well as between the two random graphs conversely. While NetSimile has a longer runtime, it also takes much more features into account. However, the Laplacian spectral distance algorithm yields similar results and might thus be a better choice when first evaluating graph distances.

[59]By comparing further museum donor networks, network similarity might be computed in a more meaningful way. As these algorithms are based on structural data, additional data might change the general shape of the networks; however, the resulting graph distances with the available data show major structural similarity between the museums. This is unlikely to change, unless a major, significantly differently structured part of the collection data has not been included in the data sets.

[60]The Laplacian spectral distance uses the lowest eigenvalue spectra from the algorithm as broadly described above for comparison between two graphs.‍[73] Lower result values indicate a smaller distance, i.e. higher similarity. Notably, the distance between the two random graphs is closer than the two museum graphs. Spectral distances, such as this one, may be used to detect both global and local features and have been reported as a good general distance measure. Our results indicate that the algorithm would be a good fit for finding network similarity.

[61]NetSimile was first proposed by Berlingerio et al. and was designed to work well on graphs of different sizes.‍[74] The algorithm uses seven different node features as detailed above which are aggregated into a single value distance afterwards. The lower the computed distance, the closer the graphs are structurally; interestingly the two lowest numbers are the comparisons between the two museums (11.741) and the two random graphs (12.185). Other comparisons yielded higher numbers, all between 22 and 23. Our results indicate that the NetSimile algorithm would also be well suited to the task presented here. A downside to this algorithm is the rather long runtime which has been noted before.‍[75]