SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1801.04062 · 2018

MINE: Mutual Information Neural Estimation

Yoshua Bengio, Aaron Courville, Devon Hjelm, Sherjil Ozair, Sai Rajeswar, Aristide Baratin, Mohamed Belghazi

arXiv · PDF · Open in the Atlas

Code that ran

9 of 24 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (4 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 7 have no contract verdict; 1 failure of our instrument; 15 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/1801.04062.json · badge: /api/ran/1801.04062.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
gtegner/mine-pytorch community
Papers with Code archive
3 ran · 3 no recorded run
mohith-sakthivel/mine-pytorch community
Papers with Code archive
3 ran · 1 no recorded run
shannonycj/simple-mine community
Papers with Code archive
1 ran · 2 no recorded run
pecholal/main-vc community
Papers with Code archive
1 ran · 0 no recorded run
copy not recorded community
not in the paper's declared repositories
1 ran · 0 no recorded run
ChengzhangZhu/MINE community
Papers with Code archive
0 ran · 3 no recorded run
personads/synvae community
Papers with Code archive
0 ran · 2 no recorded run
ReyesDeJong/MINE community
Papers with Code archive
0 ran · 2 no recorded run
wushuaida/Implementation-of-Mutual-Information-Neural-Estimator community
Papers with Code archive
0 ran · 1 no recorded run
mboudiaf/Mutual-Information-Variational-Bounds community
Papers with Code archive
0 ran · 1 no recorded run
FunctionStatusWhere it lives
ConcatLayer Ran
ran on a synthesized input
gtegner/mine-pytorch/mine/models/mine.py
code served (permissive licence) · get_code("b06f15b798bdd083")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 3, dtype family float, shape (2, 9, 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 torch
import torch.nn as nn

class ConcatLayer(nn.Module):

    def __init__(self, dim=1):
        super().__init__()
        self.dim = dim

    def forward(self, x, y):
        return torch.cat((x, y), self.dim)


INIT_ARGS = {
    'dim': 1
}

ARGS = {
    'x': torch.randn(2, 4, 3),
    'y': torch.randn(2, 5, 3)
}

if __name__ == '__main__':
    layer = ConcatLayer(**INIT_ARGS)
    output = layer(**ARGS)
    print(f"Output shape: {output.shape}")
    print(f"Output:\n{output}")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 9, 3) — shape (2, 9, 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
DistributionSimulator Ran
ran on a synthesized input
shannonycj/simple-mine/mine_estimator.py
code served (permissive licence) · get_code("95305c97fbd5a90b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned NoneType. 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

class DistributionSimulator:

    def __init__(self, x, y):
        self.x = np.array(x)
        self.y = np.array(y)
        self.standardize()

    def standardize(self):
        self.x = (self.x - self.x.mean(axis=0, keepdims=True)) / self.x.std(axis=0, keepdims=True)
        self.y = (self.y - self.y.mean()) / self.y.std()

    def reshuffle(self):
        y_cpy = self.y.copy()
        pool = np.concatenate([self.x, self.y.reshape(-1, 1)], axis=1)
        np.random.shuffle(pool)
        np.random.shuffle(y_cpy)
        self.pool = np.concatenate([pool, y_cpy.reshape(-1, 1)], axis=1)

    def init_batches(self, batch_size):
        self.batch_size = batch_size
        self.n_batch = self.x.shape[0] // batch_size + (self.x.shape[0] % batch_size > 0) * 1
        self.batch_num = 0

    def next_batch(self):
        end_idx = min((self.batch_num + 1) * self.batch_size, self.pool.shape[0])
        batch = self.pool[self.batch_size * self.batch_num:end_idx]
        self.batch_num += 1
        return (batch, end_idx == self.pool.shape[0])


INIT_ARGS = {
    'x': np.random.randn(8, 4),
    'y': np.random.randn(8)
}

ARGS = {
    'batch_size': 3
}

METHOD = 'init_batches'

if __name__ == '__main__':
    sim = DistributionSimulator(**INIT_ARGS)
    sim.reshuffle()
    sim.init_batches(**ARGS)
    batch, is_last = sim.next_batch()
    print(f"Batch shape: {batch.shape}, Is last: {is_last}")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type NoneType — a NoneType output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a NoneType 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
MINE Ran
ran on a synthesized input · constructed an object; no result was computed
pecholal/main-vc/models/mi.py
code served (permissive licence) · get_code("72dc7c30ba369ffe")
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 an instance of its own class, MINE: the run constructed the object and did not call it, so no result was computed. 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 torch
import numpy
import torch.nn as nn

torch.manual_seed(0)

class MINE(nn.Module):

    def __init__(self, x_dim, y_dim, hidden_size):
        super(MINE, self).__init__()
        self.T_func = nn.Sequential(nn.Linear(x_dim + y_dim, hidden_size), nn.ReLU(), nn.Linear(hidden_size, hidden_size), nn.ReLU(), nn.Linear(hidden_size, 1))

    def forward(self, x_samples, y_samples):
        x_samples = x_samples.unsqueeze(1).expand(-1, y_samples.shape[1], -1).reshape(-1, x_samples.shape[-1])
        y_samples = y_samples.reshape(-1, y_samples.shape[-1])
        sample_size = y_samples.shape[0]
        tiled_x = torch.cat([x_samples, x_samples], dim=0)
        random_index = torch.randperm(sample_size)
        y_shuffle = y_samples[random_index]
        concat_y = torch.cat([y_samples, y_shuffle], dim=0)
        inputs = torch.cat([tiled_x, concat_y])
        logits = self.T_func(inputs)
        tmp_1 = logits[:sample_size]
        tmp_2 = logits[sample_size:]
        lower_bound = numpy.log2(numpy.exp(1)) * (torch.mean(tmp_1) - torch.log(torch.mean(torch.exp(tmp_2))))
        return lower_bound

    def learning_loss(self, x_samples, y_samples):
        return -self.forward(x_samples, y_samples)


x_dim = 4
y_dim = 3
hidden_size = 8

x_samples = torch.randn(2, x_dim)
y_samples = torch.randn(2, 2, y_dim)

ARGS = {
    'x_dim': x_dim,
    'y_dim': y_dim,
    'hidden_size': hidden_size
}
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 MINE — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a MINE 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
MINE_Base Ran
ran on a synthesized input · constructed an object; no result was computed
mohith-sakthivel/mine-pytorch/mine/mine.py
code served (permissive licence) · get_code("cabc5db086991209")
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 an instance of its own class, MINE_Base: the run constructed the object and did not call it, so no result was computed. 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 torch
import torch.nn as nn

torch.manual_seed(0)

ARGS = {
    'input_dim': 4,
    'K': 3,
    'est_lr': 0.0002,
    'variant': 'unbiased'
}
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 MINE_Base — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a MINE_Base 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
MINE_DV Ran
ran on a synthesized input · constructed an object; no result was computed
mohith-sakthivel/mine-pytorch/mine/mine.py
code served (permissive licence) · get_code("86e5fe23b79fe923")
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 an instance of its own class, MINE_DV: the run constructed the object and did not call it, so no result was computed. 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 torch
import torch.nn as nn
import torch.optim as optim
from torch.autograd import Function
import torch.nn.functional as F
import math

torch.manual_seed(0)

class StatisticsNet(nn.Module):
    """
    Network for estimating mutual information between two random variables
    """

    def __init__(self, x_dim, z_dim):
        super().__init__()
        self._layers = nn.ModuleList()
        self._layers.append(nn.Linear(x_dim + z_dim, 512))
        self._layers.append(nn.Linear(512, 512))
        self._out_layer = nn.Linear(512, 1)
        self._initialize_weights()

    def _initialize_weights(self):
        for name, param in self._layers.named_parameters():
            if 'weight' in name:
                nn.init.kaiming_normal_(param, nonlinearity='relu')
            elif 'bias' in name:
                nn.init.zeros_(param)
        for name, param in self._out_layer.named_parameters():
            if 'weight' in name:
                nn.init.kaiming_normal_(param, nonlinearity='linear')
            elif 'bias' in name:
                nn.init.zeros_(param)

    def forward(self, x, z):
        x = torch.cat([x, z], dim=1)
        x = x + 0.3 * torch.randn_like(x)
        for hid_layer in self._layers:
            x = F.elu(hid_layer(x))
            x = x + 0.5 * torch.randn_like(x)
        return self._out_layer(x)

class UnbiasedLogMeanExp(Function):
    """
    Calculates and uses gradients with reduced bias
    """
    epsilon = 1e-06

    @staticmethod
    def forward(ctx, i, ema):
        ctx.save_for_backward(i, ema)
        mean_numel = torch.tensor(i.shape[0], dtype=torch.float)
        output = i.logsumexp(dim=0).subtract(torch.log(mean_numel))
        return output

    @staticmethod
    def backward(ctx, grad_output):
        i, ema = ctx.saved_tensors
        grad_i = grad_ema = None
        mean_numel = torch.tensor(i.shape[0], dtype=torch.float)
        grad_i = grad_output * (i.exp() / ((ema + UnbiasedLogMeanExp.epsilon) * mean_numel))
        return (grad_i, grad_ema)

class MINE_Base(nn.Module):

    def __init__(self, input_dim, K, est_lr=0.0002, variant='unbiased'):
        super().__init__()
        self._T = StatisticsNet(input_dim, K)
        self._est_lr = est_lr
        self.variant = variant
        self._current_epoch = 0

    def _configure_optimizers(self):
        opt = optim.Adam(self._T.parameters(), lr=self._est_lr, betas=(0.5, 0.999))
        sch = None
        return (opt, sch)

    def step_epoch(self):
        self._current_epoch += 1

class MINE_DV(MINE_Base):
    _ANNEAL_PERIOD = 0
    _EMA_ANNEAL_PERIOD = 0

    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        self._decay = 0.994
        self._ema = None

    def _update_ema(self, t_margin):
        with torch.no_grad():
            exp_t = t_margin.exp().mean(dim=0)
            if self._ema is not None:
                self._ema = self._decay * self._ema + (1 - self._decay) * exp_t
            else:
                self._ema = exp_t

    def get_mi_bound(self, x, z, z_margin=None, update_ema=False):
        t_joint = self._T(x, z).mean(dim=0)
        if z_margin is not None:
            t_margin = self._T(x, z_margin)
        else:
            t_margin = self._T(x, z[torch.randperm(x.shape[0])])
        if (self.variant == 'unbiased' and update_ema) and self._current_epoch > self._EMA_ANNEAL_PERIOD:
            self._update_ema(t_margin)
        if self.variant == 'unbiased' and self._current_epoch > self._ANNEAL_PERIOD:
            log_exp_t = UnbiasedLogMeanExp.apply(t_margin, self._ema)
        else:
            log_exp_t = t_margin.logsumexp(dim=0).subtract(math.log(x.shape[0]))
        return t_joint - log_exp_t

ARGS = {
    'input_dim': 4,
    'K': 3,
    'est_lr': 0.0002,
    'variant': 'unbiased',
}
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 MINE_DV — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a MINE_DV 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
StatisticsNet Ran
ran on a synthesized input · constructed an object; no result was computed
mohith-sakthivel/mine-pytorch/mine/mine.py
code served (permissive licence) · get_code("7716a6d738daf434")
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 an instance of its own class, StatisticsNet: the run constructed the object and did not call it, so no result was computed. 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 torch
import torch.nn as nn

torch.manual_seed(0)

# StatisticsNet.__init__ takes x_dim and z_dim
# These should be distinct integers representing input dimensions
x_dim = 4
z_dim = 3

ARGS = {
    'x_dim': x_dim,
    'z_dim': z_dim,
}
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 StatisticsNet — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a StatisticsNet 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
T Ran
ran on a synthesized input
gtegner/mine-pytorch/mine/models/mine.py
code served (permissive licence) · get_code("9344ddbf1a3fe029")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 2, dtype family float, shape (2, 1) 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 torch.nn as nn

class ConcatLayer(nn.Module):

    def __init__(self, dim=1):
        super().__init__()
        self.dim = dim

    def forward(self, x, y):
        return torch.cat((x, y), self.dim)

class CustomSequential(nn.Sequential):

    def forward(self, *input):
        for module in self._modules.values():
            if isinstance(input, tuple):
                input = module(*input)
            else:
                input = module(input)
        return input

class T(nn.Module):

    def __init__(self, x_dim, z_dim):
        super().__init__()
        self.layers = CustomSequential(ConcatLayer(), nn.Linear(x_dim + z_dim, 400), nn.ReLU(), nn.Linear(400, 400), nn.ReLU(), nn.Linear(400, 400), nn.ReLU(), nn.Linear(400, 1))

    def forward(self, x, z):
        return self.layers(x, z)


INIT_ARGS = {
    'x_dim': 4,
    'z_dim': 3,
}

ARGS = {
    'x': torch.randn(2, 4),
    'z': torch.randn(2, 3),
}

model = T(**INIT_ARGS)
output = model(**ARGS)
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 1) — shape (2, 1) 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
ema Ran
ran, honoured the contract we drafted
this paper's copy was not recorded; identical code first harvested from gtegner/mine-pytorch
pointer only · get_code("a1073d8f05df7b5a")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

ema_loss Ran
ran; our contract draft was wrong, not the code
gtegner/mine-pytorch/mine/models/mine.py
code served (permissive licence) · get_code("ea691d8a19584637")
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 torch
import math

torch.manual_seed(0)

# Create a small input tensor (batch_size=2, dim=4)
x = torch.randn(2, 4, requires_grad=True)

# Initialize running_mean as 0 (will be set to t_exp on first call)
running_mean = 0

# Set alpha parameter for EMA
alpha = 0.9

ARGS = {
    "x": x,
    "running_mean": running_mean,
    "alpha": alpha,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — 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
BaseModel Not yet run
unverified (harvested, no recorded run)
personads/synvae/models/mine.py
code served (permissive licence) · get_code("c2fcf423ea534add")
How it ran

no invocation recorded.

CustomSequential Not yet run
unverified (harvested, no recorded run)
gtegner/mine-pytorch/mine/models/mine.py
code served (permissive licence) · get_code("1db351adf24cdd54")
How it ran

no invocation recorded.

EMALoss Not yet run
unverified (harvested, no recorded run)
gtegner/mine-pytorch/mine/models/mine.py
code served (permissive licence) · get_code("9fa9099208a14812")
How it ran

no invocation recorded.

GradMINE Not yet run
unverified (harvested, no recorded run)
ReyesDeJong/MINE/simple_mine_f.py
code served (permissive licence) · get_code("08c96aa7e79f4bd5")
How it ran

no invocation recorded.

MINE Not yet run
unverified (harvested, no recorded run)
wushuaida/Implementation-of-Mutual-Information-Neural-Estimator/MINE.py
pointer only (licence: NONE) · get_code("5d49c337f7803d05")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, wushuaida/Implementation-of-Mutual-Information-Neural-Estimator when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

MINE Not yet run
unverified (harvested, no recorded run)
ChengzhangZhu/MINE/mine.py
pointer only (licence: NONE) · get_code("4e9109b79db8ab70")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, ChengzhangZhu/MINE when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

MINEf Not yet run
unverified (harvested, no recorded run)
ReyesDeJong/MINE/simple_mine_f.py
code served (permissive licence) · get_code("c2fa6bf887beac14")
How it ran

no invocation recorded.

Mi_estimator Not yet run
unverified (harvested, no recorded run)
mboudiaf/Mutual-Information-Variational-Bounds/Mi_estimator.py
pointer only (licence: NONE) · get_code("3dc3cc33422612a7")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, mboudiaf/Mutual-Information-Variational-Bounds when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Mine Not yet run
unverified (harvested, no recorded run)
personads/synvae/models/mine.py
code served (permissive licence) · get_code("49ff1c351499b6ee")
How it ran

no invocation recorded.

Mine Not yet run
unverified (harvested, no recorded run)
gtegner/mine-pytorch/mine/models/mine.py
code served (permissive licence) · get_code("d9c957348b294890")
How it ran

no invocation recorded.

UnbiasedLogMeanExp Not yet run
unverified (harvested, no recorded run)
mohith-sakthivel/mine-pytorch/mine/mine.py
code served (permissive licence) · get_code("baf7cd49a939dd37")
How it ran

no invocation recorded.

build_net Not yet run
unverified (harvested, no recorded run)
shannonycj/simple-mine/mine_estimator.py
code served (permissive licence) · get_code("b40580a228020fe9")
How it ran

no invocation recorded.

mine Not yet run
unverified (harvested, no recorded run)
shannonycj/simple-mine/mine_estimator.py
code served (permissive licence) · get_code("352ceb024666d2cd")
How it ran

no invocation recorded.

mine_loss Not yet run
unverified (harvested, no recorded run)
ChengzhangZhu/MINE/mine.py
pointer only (licence: NONE) · get_code("3f475d5fa5c5796c")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, ChengzhangZhu/MINE when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

shuffle Not yet run
unverified (harvested, no recorded run)
ChengzhangZhu/MINE/mine.py
pointer only (licence: NONE) · get_code("7697af4755f25282")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, ChengzhangZhu/MINE when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Also present in other papers' code

Of 24 samples lifted from this paper's linked repositories, 24 carry a function identity and 0 do not. Of the 24 distinct functions with an identity, 6 are also present in other papers whose code we hold (6 with at least one paper dated earlier, 0 the same month, 2 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
ema
brick of this paper's assembly: helper
3 other papers
2 dated earlier, 1 later
1606.03657 (2016-06)
InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets
1703.00810 (2017-03)
Opening the Black Box of Deep Neural Networks via Information
2311.01489 (2023-11)
Invariant Causal Imitation Learning for Generalizable Policies
ema_loss
brick of this paper's assembly: loss
3 other papers
2 dated earlier, 1 later
1606.03657 (2016-06)
InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets
1703.00810 (2017-03)
Opening the Black Box of Deep Neural Networks via Information
2311.01489 (2023-11)
Invariant Causal Imitation Learning for Generalizable Policies
MINE_Base
brick of this paper's assembly: helper
1 other paper
1 dated earlier · 1 only as an assembly brick
1612.00410 (2016-12)
Deep Variational Information Bottleneck
MINE_DV
brick of this paper's assembly: entry
1 other paper
1 dated earlier
1612.00410 (2016-12)
Deep Variational Information Bottleneck
StatisticsNet
brick of this paper's assembly: model
1 other paper
1 dated earlier · 1 only as an assembly brick
1612.00410 (2016-12)
Deep Variational Information Bottleneck
UnbiasedLogMeanExp
brick of this paper's assembly: helper
1 other paper
1 dated earlier · 1 only as an assembly brick
1612.00410 (2016-12)
Deep Variational Information Bottleneck

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 argue that the estimation of mutual information between high dimensional continuous random variables can be achieved by gradient descent over neural networks. We present a Mutual Information Neural Estimator (MINE) that is linearly scalable in dimensionality as well as in sample size, trainable through back-prop, and strongly consistent. We present a handful of applications on which MINE can be used to minimize or maximize mutual information. We apply MINE to improve adversarially trained generative models. We also use MINE to implement the Information Bottleneck, applying it to supervised classification; our results demonstrate substantial improvement in flexibility and performance in these settings.

For agents

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

get_harvested_code_for_paper("1801.04062")
get_code_for_paper("1801.04062")
have("1801.04062")

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

curl https://syntology.ai/api/ran/1801.04062.json

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

[![Syntology run record](https://syntology.ai/api/ran/1801.04062.svg)](https://syntology.ai/paper/1801.04062)

Connect an agent — have() is free.