Search results

Scan over strength of residual connections

comp: HPOHyperparameter optimizationHyperparameter optimization

objective: performanceMight impact performance of model in terms of our metricsMight impact performance of model in terms of our metrics

prio: high

Status: Open.

#507 In gnn-tracking/gnn_tracking;

· klieret opened on Apr 1, 2024

Fine-tune GC with frozen OC stage

comp: trainer

effort: low

objective: performanceMight impact performance of model in terms of our metricsMight impact performance of model in terms of our metrics

prio: high

Status: Open.

#503 In gnn-tracking/gnn_tracking;

· klieret opened on Mar 11, 2024

Degree-weight messages

comp: model

objective: performanceMight impact performance of model in terms of our metricsMight impact performance of model in terms of our metrics

prio: high

Status: Open.

#502 In gnn-tracking/gnn_tracking;

· klieret opened on Mar 11, 2024

Switch to using MLPs with residual connections in INs

comp: model

objective: performanceMight impact performance of model in terms of our metricsMight impact performance of model in terms of our metrics

prio: high

Status: Open.

#501 In gnn-tracking/gnn_tracking;

· klieret opened on Mar 11, 2024

Higher GC k but removing skip connections

objective: performanceMight impact performance of model in terms of our metricsMight impact performance of model in terms of our metrics

status: idea

type: side-studyIdea that is a bit more detached and involvedIdea that is a bit more detached and involved

Status: Open.

#499 In gnn-tracking/gnn_tracking;

· klieret opened on Mar 11, 2024

Weight loss by track pt

comp: lossesLoss functionsLoss functions

effort: low

objective: performanceMight impact performance of model in terms of our metricsMight impact performance of model in terms of our metrics

prio: normal

Status: Open.

#257 In gnn-tracking/gnn_tracking;

· GageDeZoort opened on Feb 16, 2023

Investigate iterative clustering as proposed in original paper

comp: cluster

objective: performanceMight impact performance of model in terms of our metricsMight impact performance of model in terms of our metrics

prio: normal

status: idea

type: side-studyIdea that is a bit more detached and involvedIdea that is a bit more detached and involved

Status: Open.

#254 In gnn-tracking/gnn_tracking;

· klieret opened on Feb 13, 2023

Use multiple aggregations in interaction networks

comp: model

objective: performanceMight impact performance of model in terms of our metricsMight impact performance of model in terms of our metrics

prio: low

status: idea

Status: Open.

#249 In gnn-tracking/gnn_tracking;

· klieret opened on Feb 10, 2023

Do categorical EC rather than binary classification

comp: lossesLoss functionsLoss functions

comp: model

objective: performanceMight impact performance of model in terms of our metricsMight impact performance of model in terms of our metrics

status: idea

Status: Open.

#235 In gnn-tracking/gnn_tracking;

· klieret opened on Jan 27, 2023

Try gradually decaying repulsive potential rather than hinge loss

comp: lossesLoss functionsLoss functions

objective: performanceMight impact performance of model in terms of our metricsMight impact performance of model in terms of our metrics

status: idea

Status: Open.

#209 In gnn-tracking/gnn_tracking;

· klieret opened on Dec 5, 2022

Normalizations (GraphNorm, etc.)

comp: model

objective: performanceMight impact performance of model in terms of our metricsMight impact performance of model in terms of our metrics

prio: normal

status: idea

Status: Open.

#172 In gnn-tracking/gnn_tracking;

· klieret opened on Oct 26, 2022

Use attention mechanism

comp: model

objective: performanceMight impact performance of model in terms of our metricsMight impact performance of model in terms of our metrics

prio: normal

status: idea

Status: Open.

#130 In gnn-tracking/gnn_tracking;

· klieret opened on Oct 17, 2022