SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1907.10982 · 2019

Overfitting of neural nets under class imbalance: Analysis and improvements for segmentation

arXiv · PDF · Open in the Atlas

Code that ran

10 of 17 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; 0 honoured the contract we drafted, 0 violated it and 10 have no contract verdict; 0 failures of our instrument; 7 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/1907.10982.json · badge: /api/ran/1907.10982.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
ZerojumpLine/OverfittingUnderClassImbalance canonical (classifier label)
classifier basis: census
10 ran · 7 no recorded run
FunctionStatusWhere it lives
applyBn Ran
ran on a synthesized input
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/deepmedic/neuralnet/ops.py
code served (permissive licence) · get_code("63cc61fbcb26e1c1")
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 numpy as np
import tensorflow as tf

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

# Define parameters
batch_size = 2
num_channels = 4
spatial_dim_1 = 3
spatial_dim_2 = 5
spatial_dim_3 = 6

# Create input tensors with shape (batch, channels, dim1, dim2, dim3)
inputShapeTrain = (batch_size, num_channels, spatial_dim_1, spatial_dim_2, spatial_dim_3)
inputTrain = tf.constant(np.random.randn(*inputShapeTrain).astype('float32'))
inputVal = tf.constant(np.random.randn(*inputShapeTrain).astype('float32'))
inputTest = tf.constant(np.random.randn(*inputShapeTrain).astype('float32'))

# Rolling average window size
rollingAverageForBatchNormalizationOverThatManyBatches = 3

ARGS = {
    'rollingAverageForBatchNormalizationOverThatManyBatches': rollingAverageForBatchNormalizationOverThatManyBatches,
    'inputTrain': inputTrain,
    'inputVal': inputVal,
    'inputTest': inputTest,
    'inputShapeTrain': inputShapeTrain,
}
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
calcRecFieldFromKernDimListPerLayerWhenStrides1 Ran
ran on a synthesized input
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/deepmedic/neuralnet/utils.py
code served (permissive licence) · get_code("93947ffa6feaaa08")
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 list, ndim 1, dtype family int; 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

torch.manual_seed(0)

# The function takes a list of kernel dimension lists
# Each element represents the kernel dimensions for a layer across all spatial dimensions
# For example: [[3, 3], [5, 5]] means 2 layers, each with 2D kernels

kernDimPerLayerList = [
    [3, 5],      # Layer 0: kernel size 3 in dim 0, kernel size 5 in dim 1
    [3, 3],      # Layer 1: kernel size 3 in dim 0, kernel size 3 in dim 1
    [5, 3]       # Layer 2: kernel size 5 in dim 0, kernel size 3 in dim 1
]

ARGS = {
    "kernDimPerLayerList": kernDimPerLayerList
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family int — a list 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
checkRecFieldVsSegmSize Ran
ran on a synthesized input
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/deepmedic/neuralnet/utils.py
code served (permissive licence) · get_code("56a35a765c4ddb29")
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 bool, 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 torch
import numpy as np

# Call torch.manual_seed for reproducibility
torch.manual_seed(0)

# Based on the function body:
# - receptiveFieldDim: used with len() and indexed with dim_i, compared element-wise with segmentDim
# - segmentDim: used with len() and indexed with dim_i, compared element-wise with receptiveFieldDim
# Both should be sequences (list, tuple) of integers representing dimensions
# The function checks that RF dimensions don't exceed segment dimensions

# Create valid arguments where segment is larger than receptive field
receptiveFieldDim = [3, 5, 4]  # RF dimensions
segmentDim = [5, 8, 7]         # Segment dimensions (all >= RF dimensions)

ARGS = {
    "receptiveFieldDim": receptiveFieldDim,
    "segmentDim": segmentDim
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, 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
cropRczOf5DimArrayToMatchOther Ran
ran on a synthesized input
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/deepmedic/neuralnet/pathways.py
code served (permissive licence) · get_code("5caf38aae31d6629")
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 Tensor, ndim 5, dtype family float, shape (2, 3, 4, 5, 6) 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 torch
import numpy as np

torch.manual_seed(0)

# Create a 5D array to crop with shape (batch, channels, depth, height, width)
# Using distinct sizes: batch=2, channels=3, depth=5, height=6, width=7
array5DimToCrop = torch.randn(2, 3, 5, 6, 7)

# Create dimensions to match with shape (batch, channels, depth, height, width)
# Using distinct sizes to match against
dimensionsOf5DimArrayToMatchInRcz = torch.tensor([2, 3, 4, 5, 6])

ARGS = {
    "array5DimToCrop": array5DimToCrop,
    "dimensionsOf5DimArrayToMatchInRcz": dimensionsOf5DimArrayToMatchInRcz
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 5, dtype family float, shape (2, 3, 4, 5, 6) — shape (2, 3, 4, 5, 6) 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
getEpochsBetweenFullInf Ran
ran on a synthesized input
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/plotTrainingProgress.py
code served (permissive licence) · get_code("dec48a9241121f7d")
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 int, ndim 0, dtype family int, 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 tempfile
import os

# Create a temporary log file with the expected pattern
temp_dir = tempfile.mkdtemp()
log_file_path = os.path.join(temp_dir, "test_log.txt")

# Write a log file with the pattern and a value
with open(log_file_path, 'w') as f:
    f.write("Some initial log line\n")
    f.write(f"Perform Full-Inference on Val. cases every that many epochs = 5\n")
    f.write("Some other log line\n")

ARGS = {
    'pathToLog': log_file_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, 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
getNameOfLogFileWithoutEnding Ran
ran on a synthesized input
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/plotTrainingProgress.py
code served (permissive licence) · get_code("3349f980a9bcf37d")
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 str. 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

def getNameOfLogFileWithoutEnding(filePathToLog):
    filenameOfLog = os.path.basename(filePathToLog)
    filenameWithoutExt, extension1 = os.path.splitext(filenameOfLog)
    return filenameWithoutExt

ARGS = {
    "filePathToLog": "/path/to/logfile.txt"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
getSubepochsPerEpoch Ran
ran on a synthesized input
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/plotTrainingProgress.py
code served (permissive licence) · get_code("224bcebf7e0e8b89")
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 int, ndim 0, dtype family int, 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 tempfile
import os

# Create a temporary log file with the expected pattern
temp_dir = tempfile.mkdtemp()
log_file_path = os.path.join(temp_dir, "test_log.txt")

with open(log_file_path, 'w') as f:
    f.write("Some initial log line\n")
    f.write("Another log line\n")
    f.write("Number of Subepochs per epoch = 5\n")
    f.write("More log lines after\n")

ARGS = {
    'pathToLog': log_file_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, 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
get_normalized_vector Ran
ran on a synthesized input
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/deepmedic/neuralnet/cnn3d.py
code served (permissive licence) · get_code("ab8004d0b5654bc2")
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 2, dtype family float, shape (2, 4) 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

def get_normalized_vector(d):
    d /= 1e-12 + tf.reduce_max(tf.abs(d), list(range(1, len(d.get_shape()))), keepdims=True)
    d /= tf.sqrt(1e-06 + tf.reduce_sum(tf.pow(d, 2.0), list(range(1, len(d.get_shape()))), keepdims=True))
    return d

# Create a small tensor with shape (batch=2, features=4)
# The function expects a tensor with at least 2 dimensions
d = tf.constant([[1.0, 2.0, 3.0, 4.0],
                 [5.0, 6.0, 7.0, 8.0]], dtype=tf.float32)

ARGS = {
    "d": d
}
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 2, dtype family float, shape (2, 4) — shape (2, 4) 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
padImageWithMirroring Ran
ran on a synthesized input
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/deepmedic/neuralnet/cnn3d.py
code served (permissive licence) · get_code("c3e2ff49dcdf5978")
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 5, 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 numpy as np
import tensorflow as tf

# Create a small 5D input image tensor
# Shape: (batch, channel, depth, height, width)
batch_size = 2
channels = 3
depth = 4
height = 5
width = 6

inputImage = tf.constant(
    np.random.randn(batch_size, channels, depth, height, width).astype(np.float32)
)

# Create padding specification for each dimension (depth, height, width)
voxelsPerDimToPad = np.array([2, 4, 6], dtype=np.int32)

ARGS = {
    "inputImage": inputImage,
    "voxelsPerDimToPad": voxelsPerDimToPad,
}
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 5, dtype family float — no output shape is stated: observe_signature.py recorded SymbolicTensor ndim 5, which disagrees with the fixture run
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
repeatRcz5DimArrayByFactor Ran
ran on a synthesized input
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/deepmedic/neuralnet/pathways.py
code served (permissive licence) · get_code("8ccd7fca225d0c27")
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 5, dtype family float, shape (2, 3, 8, 15, 12) 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 5D array with shape [batch=2, dim1=3, dim2=4, dim3=5, dim4=6]
array5Dim = tf.constant(np.random.randn(2, 3, 4, 5, 6), dtype=tf.float32)

# Shape specification for the array (must match array5Dim's shape)
array5dimToUpsampleShape = [2, 3, 4, 5, 6]

# Upsampling factor for dimensions 2, 3, 4
factor3Dim = [2, 3, 2]

ARGS = {
    "array5Dim": array5Dim,
    "array5dimToUpsampleShape": array5dimToUpsampleShape,
    "factor3Dim": factor3Dim
}
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 5, dtype family float, shape (2, 3, 8, 15, 12) — shape (2, 3, 8, 15, 12) 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
applyDropout Not yet run
unverified (harvested, no recorded run)
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/deepmedic/neuralnet/ops.py
code served (permissive licence) · get_code("c71bea9d050b4a3c")
How it ran

no invocation recorded.

calculateSubsampledImagePartDimensionsFromImagePartSizePatchSizeAndSubsampleFactor Not yet run
unverified (harvested, no recorded run)
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/deepmedic/neuralnet/utils.py
code served (permissive licence) · get_code("280beba3466e6865")
How it ran

no invocation recorded.

dsc Not yet run
unverified (harvested, no recorded run)
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/deepmedic/neuralnet/cost_functions.py
code served (permissive licence) · get_code("9d5e70582a1d30c1")
How it ran

no invocation recorded.

iou Not yet run
unverified (harvested, no recorded run)
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/deepmedic/neuralnet/cost_functions.py
code served (permissive licence) · get_code("730ced37040cedf6")
How it ran

no invocation recorded.

makeBiasParamsAndApplyToFms Not yet run
unverified (harvested, no recorded run)
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/deepmedic/neuralnet/ops.py
code served (permissive licence) · get_code("7857723aa3534d8e")
How it ran

no invocation recorded.

upsampleRcz5DimArrayAndOptionalCrop Not yet run
unverified (harvested, no recorded run)
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/deepmedic/neuralnet/pathways.py
code served (permissive licence) · get_code("c470664471f70591")
How it ran

no invocation recorded.

x_entr Not yet run
unverified (harvested, no recorded run)
ZerojumpLine/OverfittingUnderClassImbalance/DeepMedic/deepmedic/neuralnet/cost_functions.py
code served (permissive licence) · get_code("343c57cfda89db25")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 0 later) and 15 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
dsc 1 other paper
1 dated earlier
1603.05959 (2016-03)
Efficient Multi-Scale 3D CNN with Fully Connected CRF for Accurate Brain Lesion Segmentation
iou 1 other paper
1 dated earlier
1603.05959 (2016-03)
Efficient Multi-Scale 3D CNN with Fully Connected CRF for Accurate Brain Lesion Segmentation

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

Overfitting in deep learning has been the focus of a number of recent works, yet its exact impact on the behavior of neural networks is not well understood. This study analyzes overfitting by examining how the distribution of logits alters in relation to how much the model overfits. Specifically, we find that when training with few data samples, the distribution of logit activations when processing unseen test samples of an under-represented class tends to shift towards and even across the decision boundary, while the over-represented class seems unaffected. In image segmentation, foreground samples are often heavily under-represented. We observe that sensitivity of the model drops as a result of overfitting, while precision remains mostly stable. Based on our analysis, we derive asymmetric modifications of existing loss functions and regularizers including a large margin loss, focal loss, adversarial training and mixup, which specifically aim at reducing the shift observed when embedding unseen samples of the under-represented class. We study the case of binary segmentation of brain tumor core and show that our proposed simple modifications lead to significantly improved segmentation performance over the symmetric variants.

For agents

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

get_harvested_code_for_paper("1907.10982")
get_code_for_paper("1907.10982")
have("1907.10982")

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

curl https://syntology.ai/api/ran/1907.10982.json

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

[![Syntology run record](https://syntology.ai/api/ran/1907.10982.svg)](https://syntology.ai/paper/1907.10982)

Connect an agent — have() is free.