SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2201.12204 · 2022

From data to functa: Your data point is a function and you can treat it like one

arXiv · PDF · Open in the Atlas

Code that ran

9 of 11 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 9 have no contract verdict; 0 failures of our instrument; 2 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/2201.12204.json · badge: /api/ran/2201.12204.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
deepmind/functa canonical (classifier label)
classifier basis not recorded on the link
9 ran · 2 no recorded run
FunctionStatusWhere it lives
array_to_flattened_pytree Ran
ran on a synthesized input
deepmind/functa/pytree_conversions.py
code served (permissive licence) · get_code("49ed47f8d0f2820c")
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. 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

# Construct test data for array_to_flattened_pytree
# The function takes a concatenated array and indices that mark boundaries between arrays
# It reconstructs the original list of arrays by slicing at those indices

# Create a concatenated array with some sample data
concat_array = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10], dtype=np.float32)

# Create indices that mark where each original array ends
# For example: [3, 6, 10] means:
# - First array: indices 0:3 -> [1, 2, 3]
# - Second array: indices 3:6 -> [4, 5, 6]
# - Third array: indices 6:10 -> [7, 8, 9, 10]
concat_idx = [3, 6, 10]

ARGS = {
    "concat_array": concat_array,
    "concat_idx": concat_idx,
}
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 — a list 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
filter_srn Ran
ran on a synthesized input
deepmind/functa/data_utils.py
code served (permissive licence) · get_code("3b6360dafe2a0550")
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.compat.v2 as tf
import numpy as np

# Create a small batch with images
# The function expects batch['images'] with shape (batch_size, height, width, channels)
# We use distinct sizes: batch=2, height=4, width=3, channels=5

batch_size = 2
height = 4
width = 3
channels = 5

# Create image data with varying values so min_diff is non-zero
images_data = np.array([
    # First image: values from 1 to 10
    np.arange(height * width * channels, dtype=np.float32).reshape(height, width, channels) + 1,
    # Second image: values from 10 to 20
    np.arange(height * width * channels, dtype=np.float32).reshape(height, width, channels) + 10
])

images_tensor = tf.constant(images_data, dtype=tf.float32)

batch = {
    'images': images_tensor
}

ARGS = {
    'batch': batch
}
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
flattened_pytree_to_array Ran
ran on a synthesized input
deepmind/functa/pytree_conversions.py
code served (permissive licence) · get_code("f021d7c7c5e9bbc2")
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 jax
import jax.numpy as jnp
import numpy as np

# Construct a flattened pytree - a list of 1D arrays
# The function expects this to be the output of jax.tree_flatten
flattened_pytree = [
    jnp.array([1.0, 2.0, 3.0]),
    jnp.array([4.0, 5.0]),
    jnp.array([6.0, 7.0, 8.0, 9.0])
]

ARGS = {
    "flattened_pytree": flattened_pytree
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, numpy — 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
get_num_weights_and_modulations Ran
ran on a synthesized input
deepmind/functa/function_reps.py
code served (permissive licence) · get_code("d833751b40e23fc3")
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 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 haiku as hk
import jax
import jax.numpy as jnp

# Create a minimal ModulatedSiren-like parameter structure
# We need to construct hk.Params which is a nested dict structure

def create_sample_params():
    """Create a sample parameter structure mimicking ModulatedSiren."""
    params = {
        'layer_1': {
            'w': jnp.ones((4, 3)),
            'b': jnp.ones((3,)),
        },
        'fi_lm_1': {
            'scale': jnp.ones((3, 2)),
            'shift': jnp.ones((3, 2)),
        },
        'layer_2': {
            'w': jnp.ones((3, 5)),
            'b': jnp.ones((5,)),
        },
        'fi_lm_2': {
            'scale': jnp.ones((5, 2)),
            'shift': jnp.ones((5, 2)),
        },
        'latent_vector': {
            'embedding': jnp.ones((8,)),
        },
    }
    return params

params = create_sample_params()

ARGS = {
    'params': params,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
haiku, jax — 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 int — 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
get_rays Ran
ran on a synthesized input
deepmind/functa/minimal_nerf.py
code served (permissive licence) · get_code("7d26fe749fa567f8")
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 ArrayImpl, ndim 4, dtype family float, shape (2, 2, 3, 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 jax.numpy as jnp

# Set up small test values for the NeRF ray generation function
height = 2
width = 3
focal = 50.0

# Create a 4x4 pose matrix (camera to world transformation)
# Identity matrix with slight translation
pose = jnp.array([
    [1.0, 0.0, 0.0, 0.5],
    [0.0, 1.0, 0.0, 0.3],
    [0.0, 0.0, 1.0, 2.0],
    [0.0, 0.0, 0.0, 1.0]
])

ARGS = {
    "height": height,
    "width": width,
    "focal": focal,
    "pose": pose
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 4, dtype family float, shape (2, 2, 3, 3) — shape (2, 2, 3, 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
partition_params Ran
ran on a synthesized input
deepmind/functa/function_reps.py
code served (permissive licence) · get_code("60304ec2c3729e2c")
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 haiku as hk
import jax
import jax.numpy as jnp

# Create a minimal ModulatedSiren-like parameter structure
# The function uses hk.data_structures.partition which expects hk.Params (nested dicts)

def create_dummy_params():
    """Create a minimal hk.Params structure with ModulatedSiren-like keys."""
    # hk.Params is a nested dict with structure: {module_name: {param_name: value}}
    # We need to include some parameters with 'fi_lm' and 'latent_vector' in module names
    
    params = {
        'modulated_siren/fi_lm_0': {
            'w': jnp.ones((4, 3)),
            'b': jnp.ones(3),
        },
        'modulated_siren/fi_lm_1': {
            'w': jnp.ones((3, 2)),
            'b': jnp.ones(2),
        },
        'modulated_siren/latent_vector': {
            'z': jnp.ones(8),
        },
        'modulated_siren/dense_0': {
            'w': jnp.ones((4, 5)),
            'b': jnp.ones(5),
        },
        'modulated_siren/dense_1': {
            'w': jnp.ones((5, 3)),
            'b': jnp.ones(3),
        },
    }
    
    return params

dummy_params = create_dummy_params()

ARGS = {
    'params': dummy_params,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
haiku, jax — 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
partition_shared_params Ran
ran on a synthesized input
deepmind/functa/function_reps.py
code served (permissive licence) · get_code("594fe089cd4ab588")
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 haiku as hk
import jax
import jax.numpy as jnp

# Create a minimal shared_params structure that mimics what ModulatedSiren 
# or LatentModulatedSiren would produce. The function partitions based on 
# 'meta_sgd_lrs' in the module name.

# Build a nested dict structure similar to hk.Params
shared_params = {
    'linear_layer': {
        'w': jnp.array([[1.0, 2.0], [3.0, 4.0]]),
        'b': jnp.array([0.1, 0.2]),
    },
    'meta_sgd_lrs': {
        'linear_layer': {
            'w_lr': jnp.array([0.01, 0.01]),
            'b_lr': jnp.array([0.001]),
        }
    },
    'modulation': {
        'scale': jnp.array([1.5, 1.5]),
    },
    'meta_sgd_lrs_aux': {
        'modulation': {
            'scale_lr': jnp.array([0.002, 0.002]),
        }
    }
}

ARGS = {
    'shared_params': shared_params
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
haiku, jax — 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
process_srn Ran
ran on a synthesized input
deepmind/functa/data_utils.py
code served (permissive licence) · get_code("5b8a4b0088fc2d36")
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 numpy as np
import tensorflow.compat.v2 as tf

# Create a batch dict with 'images' key containing numpy array
# The function casts it to float32 and divides by 255.0
batch = {
    'images': np.random.randint(0, 256, size=(2, 4, 3, 3), dtype=np.uint8),
}

ARGS = {
    'batch': batch
}
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 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
pytree_to_array Ran
ran on a synthesized input
deepmind/functa/pytree_conversions.py
code served (permissive licence) · get_code("254419e6e5120479")
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 jax
import jax.numpy as jnp

# Create a simple pytree structure with 1D arrays
# The function expects a pytree that can be flattened into 1D arrays

pytree = {
    'layer1': jnp.array([1.0, 2.0, 3.0, 4.0]),
    'layer2': jnp.array([5.0, 6.0]),
    'nested': {
        'sublayer': jnp.array([7.0, 8.0, 9.0])
    }
}

ARGS = {
    'pytree': pytree
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — 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
render_rays Not yet run
unverified (harvested, no recorded run)
deepmind/functa/minimal_nerf.py
code served (permissive licence) · get_code("59091fdb10977671")
How it ran

no invocation recorded.

volumetric_rendering Not yet run
unverified (harvested, no recorded run)
deepmind/functa/minimal_nerf.py
code served (permissive licence) · get_code("7ccd4109a7f17823")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 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 11 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

It is common practice in deep learning to represent a measurement of the world on a discrete grid, e.g. a 2D grid of pixels. However, the underlying signal represented by these measurements is often continuous, e.g. the scene depicted in an image. A powerful continuous alternative is then to represent these measurements using an implicit neural representation, a neural function trained to output the appropriate measurement value for any input spatial location. In this paper, we take this idea to its next level: what would it take to perform deep learning on these functions instead, treating them as data? In this context we refer to the data as functa, and propose a framework for deep learning on functa. This view presents a number of challenges around efficient conversion from data to functa, compact representation of functa, and effectively solving downstream tasks on functa. We outline a recipe to overcome these challenges and apply it to a wide range of data modalities including images, 3D shapes, neural radiance fields (NeRF) and data on manifolds. We demonstrate that this approach has various compelling properties across data modalities, in particular on the canonical tasks of generative modeling, data imputation, novel view synthesis and classification. Code: https://github.com/deepmind/functa

For agents

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

get_harvested_code_for_paper("2201.12204")
get_code_for_paper("2201.12204")
have("2201.12204")

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

curl https://syntology.ai/api/ran/2201.12204.json

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

[![Syntology run record](https://syntology.ai/api/ran/2201.12204.svg)](https://syntology.ai/paper/2201.12204)

Connect an agent — have() is free.