SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1806.10779 · 2018

Differentiable Learning-to-Normalize via Switchable Normalization

arXiv · PDF · Open in the Atlas

Code that ran

10 of 13 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; 3 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/1806.10779.json · badge: /api/ran/1806.10779.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
diogorpeixoto/SN_boxy community (classifier label)
classifier basis not recorded on the link
10 ran · 3 no recorded run
FunctionStatusWhere it lives
IOU Ran
ran on a synthesized input
diogorpeixoto/SN_boxy/deps/keras_yolov3/gen_anchors.py
code served (permissive licence) · get_code("49ed19748e06f453")
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 1, dtype family float, shape (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 numpy as np

# ann is a tuple of (width, height)
ann = (10, 8)

# centroids is an iterable of tuples, each with (centroid_width, centroid_height)
centroids = [
    (12, 10),
    (8, 6),
    (15, 5),
    (9, 9)
]

ARGS = {
    "ann": ann,
    "centroids": centroids
}
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 1, dtype family float, shape (4,) — shape (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
avg_IOU Ran
ran on a synthesized input
diogorpeixoto/SN_boxy/deps/keras_yolov3/gen_anchors.py
code served (permissive licence) · get_code("fccc5b5bb1fb6cca")
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 float32, 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 numpy as np

# Construct valid arguments for avg_IOU function
# The function expects:
# - anns: numpy array of shape (n, 2) containing annotation dimensions (w, h)
# - centroids: list/iterable of centroid tuples (c_w, c_h)

# Create a small annotation array with shape (2, 2) - 2 annotations, each with (w, h)
anns = np.array([
    [10.0, 15.0],  # annotation 1: width=10, height=15
    [20.0, 25.0],  # annotation 2: width=20, height=25
], dtype=np.float32)

# Create a list of centroids
centroids = [
    (12.0, 18.0),  # centroid 1
    (22.0, 28.0),  # centroid 2
    (8.0, 10.0),   # centroid 3
]

ARGS = {
    "anns": anns,
    "centroids": centroids,
}
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 float32, 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
bbox_iou Ran
ran on a synthesized input
diogorpeixoto/SN_boxy/deps/keras_yolov3/yolo3_one_file_to_detect_them_all.py
code served (permissive licence) · get_code("780dc46447422dd0")
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 float, 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.

class BBox:
    def __init__(self, xmin, ymin, xmax, ymax):
        self.xmin = xmin
        self.ymin = ymin
        self.xmax = xmax
        self.ymax = ymax

box1 = BBox(xmin=1.0, ymin=2.0, xmax=5.0, ymax=6.0)
box2 = BBox(xmin=3.0, ymin=1.0, xmax=7.0, ymax=5.0)

ARGS = {
    'box1': box1,
    'box2': box2,
}
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 float, 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
decode_netout Ran
ran on a synthesized input
diogorpeixoto/SN_boxy/deps/keras_yolov3/yolo3_one_file_to_detect_them_all.py
code served (permissive licence) · get_code("2d5961845a8f0c6c")
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 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

# Construct valid arguments for decode_netout function

# Parameters for the function:
# netout: network output tensor with shape (grid_h, grid_w, nb_box * (5 + nb_class))
# anchors: list of anchor widths and heights
# obj_thresh: objectness threshold
# nms_thresh: NMS threshold
# net_h: network height
# net_w: network width

grid_h = 4
grid_w = 5
nb_box = 3
nb_class = 2

# Create netout: shape (grid_h, grid_w, nb_box, 5 + nb_class)
# Then reshape to (grid_h, grid_w, nb_box * (5 + nb_class))
netout = np.random.randn(grid_h, grid_w, nb_box, 5 + nb_class).astype(np.float32)
netout = netout.reshape((grid_h, grid_w, nb_box * (5 + nb_class)))

# Create anchors: list of width/height pairs for each box
# nb_box boxes means we need 2 * nb_box anchor values
anchors = [0.5, 0.6, 0.7, 0.8, 0.9, 1.0]

obj_thresh = 0.5
nms_thresh = 0.5
net_h = 416
net_w = 416

ARGS = {
    "netout": netout,
    "anchors": anchors,
    "obj_thresh": obj_thresh,
    "nms_thresh": nms_thresh,
    "net_h": net_h,
    "net_w": net_w
}
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 list, ndim 1, dtype family float — 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
dummy_loss Ran
ran on a synthesized input
diogorpeixoto/SN_boxy/deps/keras_yolov3/yolo.py
code served (permissive licence) · get_code("3d727d3a4476f75b")
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

def dummy_loss(y_true, y_pred):
    return tf.sqrt(tf.reduce_sum(y_pred))

# Create small TensorFlow tensors for the function parameters
y_true = tf.constant([[1.0, 2.0, 3.0, 4.0],
                       [5.0, 6.0, 7.0, 8.0]], dtype=tf.float32)

y_pred = tf.constant([[0.1, 0.2, 0.3, 0.4],
                       [0.5, 0.6, 0.7, 0.8]], dtype=tf.float32)

ARGS = {
    'y_true': y_true,
    'y_pred': y_pred
}
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
get_color Ran
ran on a synthesized input
diogorpeixoto/SN_boxy/deps/keras_yolov3/utils/colors.py
code served (permissive licence) · get_code("c539797d549ec4a4")
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.

# Simple setup for get_color function
ARGS = {
    'label': 5
}
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 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
parse_json_annotations Ran
ran on a synthesized input
diogorpeixoto/SN_boxy/utils/jsonToVOC2.py
code served (permissive licence) · get_code("1cf47a59d92760fd")
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 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 json
import tempfile
from pathlib import Path

# Create a temporary JSON file with valid annotation structure
temp_dir = tempfile.TemporaryDirectory()
temp_path = Path(temp_dir.name) / "annotations.json"

# Create sample annotation data matching the function's expected structure
annotations_data = {
    "image1.jpg": {
        "vehicles": [
            {
                "AABB": {
                    "x1": 10.5,
                    "x2": 100.5,
                    "y1": 20.5,
                    "y2": 150.5
                }
            },
            {
                "AABB": {
                    "x1": 200.0,
                    "x2": 300.0,
                    "y1": 50.0,
                    "y2": 200.0
                }
            }
        ]
    },
    "image2.jpg": {
        "vehicles": [
            {
                "AABB": {
                    "x1": 5.0,
                    "x2": 95.0,
                    "y1": 15.0,
                    "y2": 145.0
                }
            }
        ]
    },
    "image3.jpg": {}
}

# Write the JSON file
with temp_path.open('w') as f:
    json.dump(annotations_data, f)

ARGS = {
    "path": temp_path,
    "resize_factor": 1
}
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 list, ndim 1, dtype family float — 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
parse_voc_annotation Ran
ran on a synthesized input
diogorpeixoto/SN_boxy/deps/keras_yolov3/voc.py
code served (permissive licence) · get_code("3ae6f4d0efbac7f2")
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 os
import tempfile
import xml.etree.ElementTree as ET
import pickle

# Create temporary directories for annotations and images
temp_dir = tempfile.mkdtemp()
ann_dir = os.path.join(temp_dir, 'annotations/')
img_dir = os.path.join(temp_dir, 'images/')
os.makedirs(ann_dir, exist_ok=True)
os.makedirs(img_dir, exist_ok=True)

# Create sample VOC XML annotation files
def create_sample_annotation(ann_path, filename, width, height, objects):
    """Create a sample VOC-style XML annotation file"""
    root = ET.Element('annotation')
    
    fn_elem = ET.SubElement(root, 'filename')
    fn_elem.text = filename
    
    size = ET.SubElement(root, 'size')
    w_elem = ET.SubElement(size, 'width')
    w_elem.text = str(width)
    h_elem = ET.SubElement(size, 'height')
    h_elem.text = str(height)
    
    for obj_data in objects:
        obj_elem = ET.SubElement(root, 'object')
        name_elem = ET.SubElement(obj_elem, 'name')
        name_elem.text = obj_data['name']
        
        bndbox = ET.SubElement(obj_elem, 'bndbox')
        xmin_elem = ET.SubElement(bndbox, 'xmin')
        xmin_elem.text = str(obj_data['xmin'])
        ymin_elem = ET.SubElement(bndbox, 'ymin')
        ymin_elem.text = str(obj_data['ymin'])
        xmax_elem = ET.SubElement(bndbox, 'xmax')
        xmax_elem.text = str(obj_data['xmax'])
        ymax_elem = ET.SubElement(bndbox, 'ymax')
        ymax_elem.text = str(obj_data['ymax'])
    
    tree = ET.ElementTree(root)
    tree.write(ann_path)

# Create two sample annotations
create_sample_annotation(
    os.path.join(ann_dir, 'image1.xml'),
    'image1.jpg',
    640,
    480,
    [
        {'name': 'person', 'xmin': 10, 'ymin': 20, 'xmax': 100, 'ymax': 200},
        {'name': 'car', 'xmin': 150, 'ymin': 50, 'xmax': 400, 'ymax': 350}
    ]
)

create_sample_annotation(
    os.path.join(ann_dir, 'image2.xml'),
    'image2.jpg',
    800,
    600,
    [
        {'name': 'person', 'xmin': 50, 'ymin': 100, 'xmax': 200, 'ymax': 400},
        {'name': 'dog', 'xmin': 300, 'ymin': 250, 'xmax': 500, 'ymax': 450}
    ]
)

# Define cache file path
cache_name = os.path.join(temp_dir, 'voc_cache.pkl')

ARGS = {
    'ann_dir': ann_dir,
    'img_dir': img_dir,
    'cache_name': cache_name,
    'labels': []
}
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 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
preprocess_input Ran
ran on a synthesized input
diogorpeixoto/SN_boxy/deps/keras_yolov3/yolo3_one_file_to_detect_them_all.py
code served (permissive licence) · get_code("601ca295e001cbcc")
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 4, dtype family float, shape (1, 416, 416, 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
import cv2

# Create a sample image with shape (height, width, channels)
# Using distinct dimensions: 480x640x3
image = np.random.randint(0, 256, size=(480, 640, 3), dtype=np.uint8)

# Define network dimensions
net_h = 416
net_w = 416

ARGS = {
    "image": image,
    "net_h": net_h,
    "net_w": net_w,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 4, dtype family float, shape (1, 416, 416, 3) — shape (1, 416, 416, 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
validate_path Ran
ran on a synthesized input
diogorpeixoto/SN_boxy/utils/jsonToVOC2.py
code served (permissive licence) · get_code("3bbf7b1aa89674c3")
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 PosixPath. 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 tempfile
from pathlib import Path

# Create a temporary directory and file to use for testing
temp_dir = tempfile.mkdtemp()
temp_file = Path(temp_dir) / "test_file.txt"
temp_file.write_text("test content")

ARGS = {
    "path": str(temp_file)
}
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 PosixPath — a PosixPath output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a PosixPath 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
draw_boxes Not yet run
unverified (harvested, no recorded run)
diogorpeixoto/SN_boxy/deps/keras_yolov3/utils/bbox.py
code served (permissive licence) · get_code("9a838de136e9595a")
How it ran

no invocation recorded.

multi_gpu_model Not yet run
unverified (harvested, no recorded run)
diogorpeixoto/SN_boxy/deps/keras_yolov3/utils/multi_gpu_model.py
code served (permissive licence) · get_code("d0a2afca0c48b31f")
How it ran

no invocation recorded.

run_kmeans Not yet run
unverified (harvested, no recorded run)
diogorpeixoto/SN_boxy/deps/keras_yolov3/gen_anchors.py
code served (permissive licence) · get_code("847e9e70cb1fb450")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 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 13 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

We address a learning-to-normalize problem by proposing Switchable Normalization (SN), which learns to select different normalizers for different normalization layers of a deep neural network. SN employs three distinct scopes to compute statistics (means and variances) including a channel, a layer, and a minibatch. SN switches between them by learning their importance weights in an end-to-end manner. It has several good properties. First, it adapts to various network architectures and tasks (see Fig.1). Second, it is robust to a wide range of batch sizes, maintaining high performance even when small minibatch is presented (e.g. 2 images/GPU). Third, SN does not have sensitive hyper-parameter, unlike group normalization that searches the number of groups as a hyper-parameter. Without bells and whistles, SN outperforms its counterparts on various challenging benchmarks, such as ImageNet, COCO, CityScapes, ADE20K, and Kinetics. Analyses of SN are also presented. We hope SN will help ease the usage and understand the normalization techniques in deep learning. The code of SN has been made available in https://github.com/switchablenorms/.

For agents

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

get_harvested_code_for_paper("1806.10779")
get_code_for_paper("1806.10779")
have("1806.10779")

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

curl https://syntology.ai/api/ran/1806.10779.json

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

[![Syntology run record](https://syntology.ai/api/ran/1806.10779.svg)](https://syntology.ai/paper/1806.10779)

Connect an agent — have() is free.