Datasets:
File size: 2,780 Bytes
6c1c833 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 | {
"field_id": "F010",
"field": "active learning",
"directions": [
{
"direction_id": "G0",
"name": "Acquiring usable annotation under effort and feedback constraints",
"scope": "When annotation cost, feedback quality, or response reliability is constrained, decide which partial annotations to obtain and which sources or fidelities to use, and make acquired or existing feedback usable for learning. These three branches form a union; a single work need not address all of them."
},
{
"direction_id": "G1",
"name": "Active learning with unlabeled-data-assisted training",
"scope": "With few task labels, use unlabeled data to establish an effective querying and learning process. The scope includes representation learning that supports acquisition and the use of pseudo-labels or consistency training after querying; neither is required to accompany the other."
},
{
"direction_id": "G2",
"name": "Assessment and limits of active learning",
"scope": "Establish interpretable comparisons of strategies and use empirical frameworks or theoretical results to determine when active querying is effective, what advantages it offers, or when it fails."
},
{
"direction_id": "G3",
"name": "Selecting effective queries from available data",
"scope": "Select queries that improve learning based on current data, models, or task structure. The scope encompasses representative coverage, identified distributional gaps, predictive uncertainty, batch nonredundancy, graph relationships, and learning or adapting the query policy itself."
},
{
"direction_id": "G4",
"name": "Active experimental measurement and discovery",
"scope": "When evaluations, observations, or physical actions are costly, plan new measurements to learn functions or systems, reconstruct spatial environments, or discover high-value candidates. Accurate approximation, candidate utility, and feasible measurement are parallel goals, not jointly required conditions."
},
{
"direction_id": "G5",
"name": "Acquisition over arriving data and learning histories",
"scope": "When acquisition decisions unfold over time, allocate limited queries and maintain model utility. Immediate handling of arriving samples and retaining learned knowledge across tasks, classes, or training rounds are distinct branches."
},
{
"direction_id": "G6",
"name": "Coordinated acquisition for coupled learning objectives",
"scope": "When query value is jointly determined by multiple interrelated tasks or different client models, coordinate annotation budgets and selection rather than optimizing a single undifferentiated predictive objective."
}
]
}
|