SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1801.05787 · 2018

Faster gaze prediction with dense networks and Fisher pruning

arXiv · PDF · Open in the Atlas

Code that ran

3 of 3 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 0 have no contract verdict; 3 failures of our instrument; 0 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/1801.05787.json · badge: /api/ran/1801.05787.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
EkdeepSLubana/OrthoReg extension (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
FunctionStatusWhere it lives
ResPruned Ran
ran; our contract draft was wrong, not the code
EkdeepSLubana/OrthoReg/pruner.py
code served (permissive licence) · get_code("d741770c08c6d1c1")
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 ResNet34_p. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch
import torch.nn as nn
import torch.nn.functional as F

torch.manual_seed(0)

# Define the layer_cfgs parameter for ResPruned
# The cfg dict needs keys: 'base', 'l1', 'l2', 'l2s', 'l3', 'l3s', 'l4', 'l4s'
# Each l* key should be a list of channel configurations for each block
# num_blocks = [3, 4, 6, 3], so:
# - layer1: 3 blocks, needs 3*2=6 values for mid_planes and out_planes
# - layer2: 4 blocks, needs 4*2=8 values
# - layer3: 6 blocks, needs 6*2=12 values
# - layer4: 3 blocks, needs 3*2=6 values

layer_cfgs = {
    'base': 16,
    'l1': [16, 16, 16, 16, 16, 16],
    'l2': [32, 32, 32, 32, 32, 32, 32, 32],
    'l3': [64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64],
    'l2s': 32,
    'l3s': 64,
    'l4': [128, 128, 128, 128, 128, 128],
    'l4s': 128,
}

ARGS = {
    'layer_cfgs': layer_cfgs,
}
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 ResNet34_p — a ResNet34_p output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet34_p 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
ResPruned_cifar Ran
ran; our contract draft was wrong, not the code
EkdeepSLubana/OrthoReg/pruner.py
code served (permissive licence) · get_code("daa1576df793359a")
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 ResNet56_p. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch
import torch.nn as nn
import torch.nn.functional as F

torch.manual_seed(0)

# Define the layer_cfgs parameter
# This is a dict that the ResNet56_p model expects
# Looking at the __init__, it needs: 'base', 'l1', 'l2', 'l2s', 'l3', 'l3s'
# Each l1, l2, l3 is a list of channel dimensions for the layers
# With 9 blocks per layer and 2 values per block, we need lists of length 18

layer_cfgs = {
    'base': 16,           # base channels for initial conv
    'l1': [16, 16] * 9,   # 9 blocks, 2 channels each (in, out)
    'l2': [32, 32] * 9,   # 9 blocks for layer2
    'l2s': 32,            # shortcut channels for layer2
    'l3': [64, 64] * 9,   # 9 blocks for layer3
    'l3s': 64,            # shortcut channels for layer3
}

num_classes = 100

ARGS = {
    'layer_cfgs': layer_cfgs,
    'num_classes': num_classes,
}
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 ResNet56_p — a ResNet56_p output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet56_p 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
constrain_ratios Ran
ran; our contract draft was wrong, not the code
EkdeepSLubana/OrthoReg/pruner.py
code served (permissive licence) · get_code("da059431c1f96fc2")
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 ndarray, ndim 2, dtype family float, shape (10, 3) 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 numpy as np

# Create sample data for imp_order
# imp_order should be a 2D numpy array where:
# - Column 0 contains layer IDs (lids)
# - Column 1 appears to be some other data
# - Column 2 contains importance values (which get set to 10000000.0)

# Create an array with 10 rows and 3 columns
imp_order = np.array([
    [0, 1, 0.5],
    [0, 2, 0.6],
    [0, 3, 0.7],
    [1, 4, 0.8],
    [1, 5, 0.9],
    [1, 6, 0.4],
    [2, 7, 0.3],
    [2, 8, 0.2],
    [0, 9, 0.1],
    [1, 10, 0.15],
], dtype=np.float64)

# osize should be a 1D array or similar, used to compute constraints
# Based on the code: constraints = (0.05 * osize).astype(int) + 1
# osize needs to have at least as many elements as unique layer IDs
osize = np.array([100, 120, 110], dtype=np.float64)

ARGS = {
    "imp_order": imp_order,
    "osize": osize,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (10, 3) — shape (10, 3) 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

Also present in other papers' code

Of 3 samples lifted from this paper's linked repositories, 3 carry a function identity and 0 do not. Of the 3 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 3 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.

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

Predicting human fixations from images has recently seen large improvements by leveraging deep representations which were pretrained for object recognition. However, as we show in this paper, these networks are highly overparameterized for the task of fixation prediction. We first present a simple yet principled greedy pruning method which we call Fisher pruning. Through a combination of knowledge distillation and Fisher pruning, we obtain much more runtime-efficient architectures for saliency prediction, achieving a 10x speedup for the same AUC performance as a state of the art network on the CAT2000 dataset. Speeding up single-image gaze prediction is important for many real-world applications, but it is also a crucial step in the development of video saliency models, where the amount of data to be processed is substantially larger.

For agents

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

get_harvested_code_for_paper("1801.05787")
get_code_for_paper("1801.05787")
have("1801.05787")

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

curl https://syntology.ai/api/ran/1801.05787.json

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

[![Syntology run record](https://syntology.ai/api/ran/1801.05787.svg)](https://syntology.ai/paper/1801.05787)

Connect an agent — have() is free.