Sharding
Cross-source consensus on Sharding from 1 sources and 4 claims.
1 sources · 4 claims
How it works
Benefits
Risks & contraindications
Highlighted claims
- DriftXpress can partition the positive support into shards with separate landmarks, feature maps, and summaries. — DriftXpress: Faster Drifting Models via Projected RKHS Fields
- Sharding lets memory scale with the largest active shard rather than all summary workspace at once. — DriftXpress: Faster Drifting Models via Projected RKHS Fields
- The compositional attractive mean sums shard numerators and denominators before division. — DriftXpress: Faster Drifting Models via Projected RKHS Fields
- Sharding is needed at larger class counts because class-wise landmark summaries can become memory-intensive. — DriftXpress: Faster Drifting Models via Projected RKHS Fields