This might require some additional infra setup. Run the Docker image on CloudLab and load the datasets into the load.
Run one experiment per dataset and produce a latency/throughput graph per #28. The general idea is explained here, but Akshay has some research code that can help.
Figure out how to set recall at a fixed rate. Table 3 in the ACORN paper provides parameters for 80% recall.
We do not plan on creating a 3D graph. Instead, we will just report on some specified recall rate, though we may need to just state the HNSW parameters.
This might require some additional infra setup. Run the Docker image on CloudLab and load the datasets into the load.
Run one experiment per dataset and produce a latency/throughput graph per #28. The general idea is explained here, but Akshay has some research code that can help.
Figure out how to set recall at a fixed rate. Table 3 in the ACORN paper provides parameters for 80% recall.
We do not plan on creating a 3D graph. Instead, we will just report on some specified recall rate, though we may need to just state the HNSW parameters.