SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2006.05587 · ICLR · 2021

SEQUENTIAL DENSITY RATIO ESTIMATION FOR SIMULTANEOUS OPTIMIZATION OF SPEED AND ACCURACY

Akinori Ebihara, Taiki Miyagawa, Kazuyuki Sakurai, Hitoshi Imaoka

arXiv · PDF · Open in the Atlas

Code that ran

13 of 21 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 9 have no contract verdict; 3 failures of our instrument; 8 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2006.05587.json · badge: /api/ran/2006.05587.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
TaikiMiyagawa/SPRT-TANDEM canonical (classifier label)
classifier basis: census
9 ran · 6 no recorded run
taikimiyagawa/sprt-tandem canonical (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
Akinori-F-Ebihara/LLLR_toymodel community
Papers with Code archive
2 ran · 2 no recorded run
FunctionStatusWhere it lives
CE_lite Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
Akinori-F-Ebihara/LLLR_toymodel/models/losses.py
code served (permissive licence) · get_code("99a9b0009e646b84")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned EagerTensor, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf
import numpy as np

# Create small tensors for the function
batch_size = 2
num_class = 2
logit_dim = 4

# Create logits tensor (batch_size x num_class)
logits = tf.constant([[1.0, 2.0], [0.5, 1.5]], dtype=tf.float32)

# Create labels as numpy array (batch_size,)
labels = np.array([0, 1], dtype=np.int32)

ARGS = {
    "logits": logits,
    "labels": labels,
    "num_class": 2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
LLLR_lite Ran
ran, honoured the contract we drafted
Akinori-F-Ebihara/LLLR_toymodel/models/losses.py
code served (permissive licence) · get_code("9549cebefd92ba63")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned EagerTensor, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf
import numpy as np

# Create small test tensors
batch_size = 2
num_classes = 2

# logits: shape (batch_size, 2) - the function indexes [:, 0] and [:, 1]
logits = tf.constant([
    [1.5, 2.3],
    [0.8, 1.2]
], dtype=tf.float32)

# labels: shape (batch_size,) - compared with sigmoid output
labels = tf.constant([0.0, 1.0], dtype=tf.float32)

ARGS = {
    "logits": logits,
    "labels": labels
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
binary_confmx_to_bac Ran
ran on a synthesized input
TaikiMiyagawa/SPRT-TANDEM/utils/performance_metrics.py
code served (permissive licence) · get_code("dc11f5694b7e7578")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned SymbolicTensor. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf
import numpy as np

# Construct a valid binary confusion matrix
# Format: [[TP, FN], [FP, TN]]
# For a binary classification confusion matrix
confmx = tf.constant([[10, 2], [3, 15]], dtype=tf.float32)

ARGS = {
    "confmx": confmx
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type SymbolicTensor — a SymbolicTensor output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
binary_truncated_sprt Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
taikimiyagawa/sprt-tandem/utils/performance_metrics.py
code served (permissive licence) · get_code("4ce12fe48cdf3000")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf
import numpy as np

# Set seed for reproducibility
np.random.seed(0)
tf.random.set_seed(0)

# Create small test tensors
batch_size = 2
order_sprt = 2
duration_minus_order = 3
duration = duration_minus_order + order_sprt

# logits_concat shape: (batch, duration - order_sprt, order_sprt + 1, 2)
# = (2, 3, 3, 2)
logits_concat = tf.constant(
    np.random.randn(batch_size, duration_minus_order, order_sprt + 1, 2).astype(np.float32)
)

# labels_concat shape: (batch,)
# Binary labels (0 or 1)
labels_concat = tf.constant([0, 1], dtype=tf.int32)

# Parameters for SPRT
alpha = 0.05  # Type I error rate
beta = 0.05   # Type II error rate

ARGS = {
    "logits_concat": logits_concat,
    "labels_concat": labels_concat,
    "alpha": alpha,
    "beta": beta,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
calc_binary_llrs Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
taikimiyagawa/sprt-tandem/utils/performance_metrics.py
code served (permissive licence) · get_code("37d3cf0b1afae8c1")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned EagerTensor, ndim 2, dtype family float, shape (2, 7) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf
import numpy as np

# Create a small valid logits_concat tensor
# Shape: (batch, (duration - order_sprt), order_sprt + 1, 2)
# Let's use: batch=2, order_sprt=3, duration=7
# So shape should be (2, 4, 4, 2) where 4 = 7 - 3

batch_size = 2
order_sprt = 3
duration = 7
seq_length = duration - order_sprt  # 4

# Create random logits with shape (batch, seq_length, order_sprt + 1, 2)
logits_data = np.random.randn(batch_size, seq_length, order_sprt + 1, 2).astype(np.float32)
logits_concat = tf.constant(logits_data)

ARGS = {
    "logits_concat": logits_concat
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 2, dtype family float, shape (2, 7) — shape (2, 7) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
checkpoint_logger Ran
ran on a synthesized input
TaikiMiyagawa/SPRT-TANDEM/utils/util_ckpt.py
code served (permissive licence) · get_code("96672439f1928178")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import shutil
import tempfile
import tensorflow as tf

# Create a temporary directory for checkpoints
temp_dir = tempfile.mkdtemp()

# Create a simple Keras model
model = tf.keras.Sequential([
    tf.keras.layers.Dense(8, activation='relu', input_shape=(4,)),
    tf.keras.layers.Dense(2)
])

# Create an optimizer
optimizer = tf.keras.optimizers.Adam(learning_rate=0.01)

# Create a global_step variable with dtype=tf.int32
global_step = tf.Variable(0, dtype=tf.int32)

# Create a config file for testing
config_file = os.path.join(temp_dir, 'test_config.yaml')
with open(config_file, 'w') as f:
    f.write('learning_rate: 0.01\n')

# Create a checkpoint directory to resume from (for testing flag_resume=True)
resume_dir = os.path.join(temp_dir, 'resume_ckpts')
os.makedirs(resume_dir, exist_ok=True)

# Pre-populate the resume directory with a checkpoint
ckpt_for_resume = tf.train.Checkpoint(step=global_step, optimizer=optimizer, net=model)
ckpt_manager_temp = tf.train.CheckpointManager(ckpt_for_resume, resume_dir, max_to_keep=1)
ckpt_manager_temp.save()

ARGS = {
    'global_step': global_step,
    'model': model,
    'optimizer': optimizer,
    'flag_resume': True,
    'root_ckptlogs': temp_dir,
    'subproject_name': 'test_project',
    'exp_phase': 'phase1',
    'comment': 'test_run',
    'time_stamp': '20240101_120000',
    'path_resume': resume_dir,
    'max_to_keep': 3,
    'config_path': config_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
decode_feat Ran
ran on a synthesized input
TaikiMiyagawa/SPRT-TANDEM/datasets/data_processing.py
code served (permissive licence) · get_code("c3e48b4df760dba8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf
import numpy as np

# Create a small batch of features as a dictionary
batch_size = 2
duration = 3
feat_dim = 4

# Create raw video data: batch_size * duration * feat_dim elements
video_data = np.random.randn(batch_size * duration * feat_dim).astype(np.float32)
video_bytes = video_data.tobytes()

# Create labels
labels = np.array([0, 1], dtype=np.int32)

# Create a features dictionary with 'video' and 'label' keys
features = {
    'video': tf.constant(video_bytes),
    'label': tf.constant(labels)
}

ARGS = {
    'features': features,
    'duration': duration,
    'feat_dim': feat_dim,
    'dtype_feat': tf.float32,
    'dtype_label': tf.int32
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
decode_nosaic_mnist Ran
ran on a synthesized input
TaikiMiyagawa/SPRT-TANDEM/datasets/data_processing.py
code served (permissive licence) · get_code("50e5cb9e9fd3ce06")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf
import numpy as np

# Create a small batch of features that decode_nosaic_mnist expects
batch_size = 2
duration = 20
height = 28
width = 28
channels = 1

# Create raw video data as uint8 bytes
# Shape should be (batch_size * duration * height * width * channels,)
video_data = np.random.randint(0, 256, size=(batch_size * duration * height * width * channels,), dtype=np.uint8)
video_bytes = video_data.tobytes()

# Create label data
label_data = np.array([0, 1], dtype=np.int32)

# Construct the features dict that the function expects
features = {
    'video': tf.constant(video_bytes),
    'label': tf.constant(label_data, dtype=tf.int32)
}

ARGS = {
    'features': features
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_yaml Ran
ran on a synthesized input
TaikiMiyagawa/SPRT-TANDEM/utils/misc.py
code served (permissive licence) · get_code("393ae5de46d6fe89")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import yaml
import tempfile

# Create a temporary YAML file for testing
temp_dir = tempfile.mkdtemp()
yaml_path = os.path.join(temp_dir, "test_config.yaml")

# Write a valid YAML file
yaml_content = {
    "model": "test_model",
    "learning_rate": 0.001,
    "batch_size": 32,
    "layers": [1, 2, 3]
}

with open(yaml_path, 'w') as f:
    yaml.dump(yaml_content, f)

# Define arguments for the function
ARGS = {
    "yaml_path": yaml_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
yaml — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
logits_to_confmx Ran
ran on a synthesized input
TaikiMiyagawa/SPRT-TANDEM/utils/performance_metrics.py
code served (permissive licence) · get_code("d451032b3c871d46")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned SymbolicTensor. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf
import numpy as np

# Set seed for reproducibility
tf.random.set_seed(0)

# Create small tensors for testing
batch_size = 2
num_classes = 4

# Create logits: shape (batch, num_classes)
# Use random logits that will produce different predictions
logits = tf.constant([
    [1.0, 2.0, 0.5, 0.1],
    [0.1, 0.5, 2.0, 1.0]
], dtype=tf.float32)

# Create labels: shape (batch,)
# Labels should be in range [0, num_classes)
labels = tf.constant([1, 2], dtype=tf.int32)

ARGS = {
    "logits": logits,
    "labels": labels
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type SymbolicTensor — a SymbolicTensor output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
multiplet_loss_func Ran
ran on a synthesized input
TaikiMiyagawa/SPRT-TANDEM/models/losses.py
code served (permissive licence) · get_code("8175594b88600b6e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned SymbolicTensor. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tensorflow as tf

# Create small tensors for testing
batch_size = 2
order_sprt = 3  # This means logits_slice.shape[1] = order_sprt + 1 = 4
num_classes = 4

# logits_slice: shape (batch_size, order_sprt + 1, num_classes) = (2, 4, 4)
logits_slice = tf.random.normal(shape=(batch_size, order_sprt + 1, num_classes))

# labels_slice: shape (batch_size,) with class indices
labels_slice = tf.constant([0, 2], dtype=tf.int64)

ARGS = {
    "logits_slice": logits_slice,
    "labels_slice": labels_slice,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type SymbolicTensor — a SymbolicTensor output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
suggest_parameters Ran
ran on a synthesized input
TaikiMiyagawa/SPRT-TANDEM/utils/util_optuna.py
code served (permissive licence) · get_code("3d5d0d12e6b23c3b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import optuna
import torch

class MockTrial:
    """A mock trial object that implements suggest_categorical."""
    def __init__(self):
        self.suggestions = {}
    
    def suggest_categorical(self, name, choices):
        """Return the first choice from the list."""
        return choices[0]

# Create mock trial
trial = MockTrial()

# Define hyperparameter candidates
list_lr = [0.001, 0.01, 0.1]
list_bs = [16, 32, 64]
list_opt = ['adam', 'sgd', 'rmsprop']
list_do = [0.1, 0.2, 0.3]
list_wi = [32, 64, 128]

ARGS = {
    'trial': trial,
    'list_lr': list_lr,
    'list_bs': list_bs,
    'list_opt': list_opt,
    'list_do': list_do,
    'list_wi': list_wi
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
optuna, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
suggest_parameters_fe Ran
ran on a synthesized input
TaikiMiyagawa/SPRT-TANDEM/utils/util_optuna.py
code served (permissive licence) · get_code("09c06838c2fd140f")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import optuna

class MockTrial:
    """Mock trial object for optuna optimization."""
    def __init__(self):
        self.suggestions = {}
    
    def suggest_categorical(self, name, choices):
        """Return the first choice from the list."""
        return choices[0]

torch.manual_seed(0)

trial = MockTrial()
list_lr = [0.001, 0.01, 0.1]
list_bs = [16, 32, 64]
list_opt = ['adam', 'sgd', 'rmsprop']
list_do = [0.0, 0.2, 0.5]
list_wd = [0.0, 1e-5, 1e-4]

ARGS = {
    'trial': trial,
    'list_lr': list_lr,
    'list_bs': list_bs,
    'list_opt': list_opt,
    'list_do': list_do,
    'list_wd': list_wd
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
optuna, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
KLIEP_lite Not yet run
unverified (harvested, no recorded run)
Akinori-F-Ebihara/LLLR_toymodel/models/losses.py
code served (permissive licence) · get_code("1b1d4bd6e98cee84")
How it ran

no invocation recorded.

binary_llr_loss_func Not yet run
unverified (harvested, no recorded run)
TaikiMiyagawa/SPRT-TANDEM/models/losses.py
code served (permissive licence) · get_code("59f9105f519377e9")
How it ran

no invocation recorded.

confmx_to_metrics Not yet run
unverified (harvested, no recorded run)
TaikiMiyagawa/SPRT-TANDEM/utils/performance_metrics.py
code served (permissive licence) · get_code("03a548ed70ba1d0c")
How it ran

no invocation recorded.

get_batch_norm_layer Not yet run
unverified (harvested, no recorded run)
TaikiMiyagawa/SPRT-TANDEM/models/backbones_fe.py
code served (permissive licence) · get_code("61d078878105d29d")
How it ran

no invocation recorded.

get_gradient_DRE Not yet run
unverified (harvested, no recorded run)
Akinori-F-Ebihara/LLLR_toymodel/models/losses.py
code served (permissive licence) · get_code("a30e73ec3ab8a38a")
How it ran

no invocation recorded.

get_ressize_dependent_params Not yet run
unverified (harvested, no recorded run)
TaikiMiyagawa/SPRT-TANDEM/models/backbones_fe.py
code served (permissive licence) · get_code("93f9562775b33b2c")
How it ran

no invocation recorded.

np_to_tfr_nosaic_mnist Not yet run
unverified (harvested, no recorded run)
TaikiMiyagawa/SPRT-TANDEM/make_nmnist.py
code served (permissive licence) · get_code("c761d6a6ea580871")
How it ran

no invocation recorded.

read_tfrecords_nosaic_mnist Not yet run
unverified (harvested, no recorded run)
TaikiMiyagawa/SPRT-TANDEM/datasets/data_processing.py
code served (permissive licence) · get_code("9797f4b100e6ec78")
How it ran

no invocation recorded.

Also present in other papers' code

Of 21 samples lifted from this paper's linked repositories, 21 carry a function identity and 0 do not. Of the 21 distinct functions with an identity, 3 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 3 later) and 18 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
get_batch_norm_layer 1 other paper
1 later
2105.13636 (2021-05)
The Power of Log-Sum-Exp: Sequential Density Ratio Matrix Estimation for Speed-Accuracy Optimization
get_ressize_dependent_params 1 other paper
1 later
2105.13636 (2021-05)
The Power of Log-Sum-Exp: Sequential Density Ratio Matrix Estimation for Speed-Accuracy Optimization
np_to_tfr_nosaic_mnist 1 other paper
1 later
2105.13636 (2021-05)
The Power of Log-Sum-Exp: Sequential Density Ratio Matrix Estimation for Speed-Accuracy Optimization

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Classifying sequential data as early and as accurately as possible is a challenging yet critical problem, especially when a sampling cost is high. One algorithm that achieves this goal is the sequential probability ratio test (SPRT), which is known as Bayes-optimal: it can keep the expected number of data samples as small as possible, given the desired error upper-bound. However, the original SPRT makes two critical assumptions that limit its application in real-world scenarios: (i) samples are independently and identically distributed, and (ii) the likelihood of the data being derived from each class can be calculated precisely. Here, we propose the SPRT-TANDEM, a deep neural network-based SPRT algorithm that overcomes the above two obstacles. The SPRT-TANDEM sequentially estimates the loglikelihood ratio of two alternative hypotheses by leveraging a novel Loss function for Log-Likelihood Ratio estimation (LLLR) while allowing correlations up to N (∈ N) preceding samples. In tests on one original and two public video databases, Nosaic MNIST, UCF101, and SiW, the SPRT-TANDEM achieves statistically significantly better classification accuracy than other baseline classifiers, with a smaller number of data samples. The code and Nosaic MNIST are publicly available at https://github.com/TaikiMiyagawa/SPRT-TANDEM.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2006.05587")
get_code_for_paper("2006.05587")
have("2006.05587")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2006.05587.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2006.05587.svg)](https://syntology.ai/paper/2006.05587)

Connect an agent — have() is free.