Datasets:

jamie33
/

924

Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
experiment: string
alphas: list<item: double>
  child 0, item: double
conformal: struct<0.1: struct<qhat: double, coverage: double, avg_set_size: double, median_set_size: double, em (... 1085 chars omitted)
  child 0, 0.1: struct<qhat: double, coverage: double, avg_set_size: double, median_set_size: double, empty_fraction (... 286 chars omitted)
      child 0, qhat: double
      child 1, coverage: double
      child 2, avg_set_size: double
      child 3, median_set_size: double
      child 4, empty_fraction: double
      child 5, chamfer: double
      child 6, size_matched: struct<probability_topk_chamfer: double, evidence_matched_chamfer: double, gain_vs_probability_topk: (... 144 chars omitted)
          child 0, probability_topk_chamfer: double
          child 1, evidence_matched_chamfer: double
          child 2, gain_vs_probability_topk: double
          child 3, gain_vs_probability_topk_ci95: list<item: double>
              child 0, item: double
          child 4, gain_vs_evidence_matched: double
          child 5, gain_vs_evidence_matched_ci95: list<item: double>
              child 0, item: double
  child 1, 0.2: struct<qhat: double, coverage: double, avg_set_size: double, median_set_size: double, empty_fraction (... 286 chars omitted)
      child 0, qhat: double
      child 1, coverage: double
      child 2, avg_set_size: double
      child 3, median_set_size: double
      child 4, empty_fraction: double
      child 5, chamfer: double
      child 6, size_matched: st
...
 double, empty_fraction: double, n: i (... 5 chars omitted)
      child 0, coverage: double
      child 1, avg_set_size: double
      child 2, median_set_size: double
      child 3, empty_fraction: double
      child 4, n: int64
  child 1, 3: struct<coverage: double, avg_set_size: double, median_set_size: double, empty_fraction: double, n: i (... 5 chars omitted)
      child 0, coverage: double
      child 1, avg_set_size: double
      child 2, median_set_size: double
      child 3, empty_fraction: double
      child 4, n: int64
  child 2, 5: struct<coverage: double, avg_set_size: double, median_set_size: double, empty_fraction: double, n: i (... 5 chars omitted)
      child 0, coverage: double
      child 1, avg_set_size: double
      child 2, median_set_size: double
      child 3, empty_fraction: double
      child 4, n: int64
  child 3, 10: struct<coverage: double, avg_set_size: double, median_set_size: double, empty_fraction: double, n: i (... 5 chars omitted)
      child 0, coverage: double
      child 1, avg_set_size: double
      child 2, median_set_size: double
      child 3, empty_fraction: double
      child 4, n: int64
  child 4, 20: struct<coverage: double, avg_set_size: double, median_set_size: double, empty_fraction: double, n: i (... 5 chars omitted)
      child 0, coverage: double
      child 1, avg_set_size: double
      child 2, median_set_size: double
      child 3, empty_fraction: double
      child 4, n: int64
pool_top1: double
source: string
host: string
to
{'host': Value('string'), 'mode': Value('string'), 'source': Value('string'), 'q_shape_subjects_objects_cats': List(Value('int64')), 'pool_top1': Value('float64'), 'pool_top5': Value('float64'), 'topk': {'1': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64')}, '3': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64')}, '5': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64')}, '10': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64')}, '20': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64')}}, 'entropy_k': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64')}, 'split_conformal': {'0.1': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'qhat': Value('float64'), 'alpha': Value('float64')}, '0.2': {'coverage': Value('float64'), 'avg_set_size': V
...
 Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'qhat': Value('float64'), 'alpha': Value('float64')}}, 'aps': {'0.1': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'qhat': Value('float64'), 'alpha': Value('float64')}, '0.2': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'qhat': Value('float64'), 'alpha': Value('float64')}, '0.3': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'qhat': Value('float64'), 'alpha': Value('float64')}}, 'grouped_conformal': {'0.1': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'alpha': Value('float64')}, '0.2': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'alpha': Value('float64')}, '0.3': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'alpha': Value('float64')}}, 'note': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 483, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2840, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              experiment: string
              alphas: list<item: double>
                child 0, item: double
              conformal: struct<0.1: struct<qhat: double, coverage: double, avg_set_size: double, median_set_size: double, em (... 1085 chars omitted)
                child 0, 0.1: struct<qhat: double, coverage: double, avg_set_size: double, median_set_size: double, empty_fraction (... 286 chars omitted)
                    child 0, qhat: double
                    child 1, coverage: double
                    child 2, avg_set_size: double
                    child 3, median_set_size: double
                    child 4, empty_fraction: double
                    child 5, chamfer: double
                    child 6, size_matched: struct<probability_topk_chamfer: double, evidence_matched_chamfer: double, gain_vs_probability_topk: (... 144 chars omitted)
                        child 0, probability_topk_chamfer: double
                        child 1, evidence_matched_chamfer: double
                        child 2, gain_vs_probability_topk: double
                        child 3, gain_vs_probability_topk_ci95: list<item: double>
                            child 0, item: double
                        child 4, gain_vs_evidence_matched: double
                        child 5, gain_vs_evidence_matched_ci95: list<item: double>
                            child 0, item: double
                child 1, 0.2: struct<qhat: double, coverage: double, avg_set_size: double, median_set_size: double, empty_fraction (... 286 chars omitted)
                    child 0, qhat: double
                    child 1, coverage: double
                    child 2, avg_set_size: double
                    child 3, median_set_size: double
                    child 4, empty_fraction: double
                    child 5, chamfer: double
                    child 6, size_matched: st
              ...
               double, empty_fraction: double, n: i (... 5 chars omitted)
                    child 0, coverage: double
                    child 1, avg_set_size: double
                    child 2, median_set_size: double
                    child 3, empty_fraction: double
                    child 4, n: int64
                child 1, 3: struct<coverage: double, avg_set_size: double, median_set_size: double, empty_fraction: double, n: i (... 5 chars omitted)
                    child 0, coverage: double
                    child 1, avg_set_size: double
                    child 2, median_set_size: double
                    child 3, empty_fraction: double
                    child 4, n: int64
                child 2, 5: struct<coverage: double, avg_set_size: double, median_set_size: double, empty_fraction: double, n: i (... 5 chars omitted)
                    child 0, coverage: double
                    child 1, avg_set_size: double
                    child 2, median_set_size: double
                    child 3, empty_fraction: double
                    child 4, n: int64
                child 3, 10: struct<coverage: double, avg_set_size: double, median_set_size: double, empty_fraction: double, n: i (... 5 chars omitted)
                    child 0, coverage: double
                    child 1, avg_set_size: double
                    child 2, median_set_size: double
                    child 3, empty_fraction: double
                    child 4, n: int64
                child 4, 20: struct<coverage: double, avg_set_size: double, median_set_size: double, empty_fraction: double, n: i (... 5 chars omitted)
                    child 0, coverage: double
                    child 1, avg_set_size: double
                    child 2, median_set_size: double
                    child 3, empty_fraction: double
                    child 4, n: int64
              pool_top1: double
              source: string
              host: string
              to
              {'host': Value('string'), 'mode': Value('string'), 'source': Value('string'), 'q_shape_subjects_objects_cats': List(Value('int64')), 'pool_top1': Value('float64'), 'pool_top5': Value('float64'), 'topk': {'1': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64')}, '3': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64')}, '5': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64')}, '10': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64')}, '20': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64')}}, 'entropy_k': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64')}, 'split_conformal': {'0.1': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'qhat': Value('float64'), 'alpha': Value('float64')}, '0.2': {'coverage': Value('float64'), 'avg_set_size': V
              ...
               Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'qhat': Value('float64'), 'alpha': Value('float64')}}, 'aps': {'0.1': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'qhat': Value('float64'), 'alpha': Value('float64')}, '0.2': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'qhat': Value('float64'), 'alpha': Value('float64')}, '0.3': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'qhat': Value('float64'), 'alpha': Value('float64')}}, 'grouped_conformal': {'0.1': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'alpha': Value('float64')}, '0.2': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'alpha': Value('float64')}, '0.3': {'coverage': Value('float64'), 'avg_set_size': Value('float64'), 'median_set_size': Value('float64'), 'empty_fraction': Value('float64'), 'n': Value('int64'), 'alpha': Value('float64')}}, 'note': Value('string')}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

jamie33/924 — BrainSets (evidence-adaptive set size)

Public companion to jamie33/923 (BrainBits-3D runbook) and jamie33/July (code + summaries). This folder is a sibling of 923, not nested under it.

Venue decision

Target: ICLR, not ICASSP.

BrainSets is a methods claim: sample-dependent set size |C_i|, grouped/stimulus-level calibration, APS vs Top-K vs entropy-K vs split conformal, on EEG category posteriors. That is ICLR-shaped.
ICASSP 2027 is already closed (paper deadline 16 Sep 2026). A 4–5 page ICASSP cut would flatten the calibration theory that is the paper.
BrainBits-3D (Access long form) stays a separate geometry-audit paper. BrainSets reuses the same posteriors, not the same manuscript.

Machine used for CPU runs in this repo

Local MacBook Pro (2024), Apple M3, 8 cores, 8 GB RAM, arm64. Enough for NumPy set-prediction on 12 × 144 × 72 posteriors. Not enough for encoder re-inference or 3D generation (see GPU_LEFTOVER.md).

Layout

Path	What
`HOW_TO_RUN.md`	CPU jobs, expected files, HF paths
`GPU_LEFTOVER.md`	GPU-only leftovers, data addresses, no secrets
`VENUE.md`	ICLR vs ICASSP rationale
`brainsets/cpu_run.py`	CPU set algorithms (no torch)
`results/cpu/`	outputs from this Mac
`results/from_existing_brainbits/`	already-computed BrainBits numbers BrainSets needs

What is already true (no new GPU)

From BrainBits-3D / ToDo2 (see results/from_existing_brainbits/):

Collective Top-1 ≈ 4.17%, Top-5 ≈ 25% (12-subject arithmetic pool, 144 test objects, 72 cats).
Split conformal (E2): α=0.1 coverage 0.9375, mean |C| ≈ 61.2/72; APS similar size.
Geometry-risk hybrid (fixed K) already exists; BrainSets asks for adaptive K.

What this CPU pass does

Runs Top-K, entropy-K, APS, and split conformal, plus a grouped (stimulus-id) conformal variant, on:

Matched simulation (mode=sim): Dirichlet-spike posteriors calibrated to the published Top-1/Top-5 rates, so the pipeline is executable on this Mac without the 144×72 JSON dump.
Real posteriors (mode=real) when posteriors.npz is present (shape q: subjects×objects×72, y: objects).

Private p0_anchor_validation_safe/sub01.json…sub12.json on jamie33/jmtseng-work-backup are the real files (~400 KB each) but cannot be downloaded while that private dataset is over the storage quota (HTTP 403). GPU leftover #0 is: copy those JSON files into a public or quota-ok dataset, then rerun mode=real.

License

MIT.

Downloads last month: 83

Total file size:

2.78 MB