CINXE.COM

<!DOCTYPE html> <html lang="en"> <head> <meta charset="utf-8"> <meta name="viewport" content="width=device-width, initial-scale=1"> <meta name="description" content="Keras documentation"> <meta name="author" content="Keras Team"> <link rel="shortcut icon" href="https://keras.io/img/favicon.ico"> <link rel="canonical" href="https://keras.io/api/layers/initializers/" />  <meta property="og:title" content="Keras documentation: Layer weight initializers"> <meta property="og:image" content="https://keras.io/img/logo-k-keras-wb.png"> <meta name="twitter:title" content="Keras documentation: Layer weight initializers"> <meta name="twitter:image" content="https://keras.io/img/k-keras-social.png"> <meta name="twitter:card" content="summary"> <title>Layer weight initializers</title>  <link href="/css/bootstrap.min.css" rel="stylesheet">  <link href="https://fonts.googleapis.com/css2?family=Open+Sans:wght@400;600;700;800&display=swap" rel="stylesheet">  <link href="/css/docs.css" rel="stylesheet"> <link href="/css/monokai.css" rel="stylesheet">  <script>(function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start': new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0], j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src= 'https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f); })(window,document,'script','dataLayer','GTM-5DNGF4N'); </script> <script> (function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){ (i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o), m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m) })(window,document,'script','https://www.google-analytics.com/analytics.js','ga'); ga('create', 'UA-175165319-128', 'auto'); ga('send', 'pageview'); </script>  <script async defer src="https://buttons.github.io/buttons.js"></script> </head> <body>  <noscript><iframe src="https://www.googletagmanager.com/ns.html?id=GTM-5DNGF4N" height="0" width="0" style="display:none;visibility:hidden"></iframe></noscript>  <div class='k-page'> <div class="k-nav" id="nav-menu"> <a href='/'><img src='/img/logo-small.png' class='logo-small' /></a> <div class="nav flex-column nav-pills" role="tablist" aria-orientation="vertical"> <a class="nav-link" href="/about/" role="tab" aria-selected="">About Keras</a> <a class="nav-link" href="/getting_started/" role="tab" aria-selected="">Getting started</a> <a class="nav-link" href="/guides/" role="tab" aria-selected="">Developer guides</a> <a class="nav-link active" href="/api/" role="tab" aria-selected="">Keras 3 API documentation</a> <a class="nav-sublink" href="/api/models/">Models API</a> <a class="nav-sublink active" href="/api/layers/">Layers API</a> <a class="nav-sublink2" href="/api/layers/base_layer/">The base Layer class</a> <a class="nav-sublink2" href="/api/layers/activations/">Layer activations</a> <a class="nav-sublink2 active" href="/api/layers/initializers/">Layer weight initializers</a> <a class="nav-sublink2" href="/api/layers/regularizers/">Layer weight regularizers</a> <a class="nav-sublink2" href="/api/layers/constraints/">Layer weight constraints</a> <a class="nav-sublink2" href="/api/layers/core_layers/">Core layers</a> <a class="nav-sublink2" href="/api/layers/convolution_layers/">Convolution layers</a> <a class="nav-sublink2" href="/api/layers/pooling_layers/">Pooling layers</a> <a class="nav-sublink2" href="/api/layers/recurrent_layers/">Recurrent layers</a> <a class="nav-sublink2" href="/api/layers/preprocessing_layers/">Preprocessing layers</a> <a class="nav-sublink2" href="/api/layers/normalization_layers/">Normalization layers</a> <a class="nav-sublink2" href="/api/layers/regularization_layers/">Regularization layers</a> <a class="nav-sublink2" href="/api/layers/attention_layers/">Attention layers</a> <a class="nav-sublink2" href="/api/layers/reshaping_layers/">Reshaping layers</a> <a class="nav-sublink2" href="/api/layers/merging_layers/">Merging layers</a> <a class="nav-sublink2" href="/api/layers/activation_layers/">Activation layers</a> <a class="nav-sublink2" href="/api/layers/backend_specific_layers/">Backend-specific layers</a> <a class="nav-sublink" href="/api/callbacks/">Callbacks API</a> <a class="nav-sublink" href="/api/ops/">Ops API</a> <a class="nav-sublink" href="/api/optimizers/">Optimizers</a> <a class="nav-sublink" href="/api/metrics/">Metrics</a> <a class="nav-sublink" href="/api/losses/">Losses</a> <a class="nav-sublink" href="/api/data_loading/">Data loading</a> <a class="nav-sublink" href="/api/datasets/">Built-in small datasets</a> <a class="nav-sublink" href="/api/applications/">Keras Applications</a> <a class="nav-sublink" href="/api/mixed_precision/">Mixed precision</a> <a class="nav-sublink" href="/api/distribution/">Multi-device distribution</a> <a class="nav-sublink" href="/api/random/">RNG API</a> <a class="nav-sublink" href="/api/utils/">Utilities</a> <a class="nav-sublink" href="/api/keras_tuner/">KerasTuner</a> <a class="nav-sublink" href="/api/keras_cv/">KerasCV</a> <a class="nav-sublink" href="/api/keras_nlp/">KerasNLP</a> <a class="nav-sublink" href="/api/keras_hub/">KerasHub</a> <a class="nav-link" href="/2.18/api/" role="tab" aria-selected="">Keras 2 API documentation</a> <a class="nav-link" href="/examples/" role="tab" aria-selected="">Code examples</a> <a class="nav-link" href="/keras_tuner/" role="tab" aria-selected="">KerasTuner: Hyperparameter Tuning</a> <a class="nav-link" href="/keras_hub/" role="tab" aria-selected="">KerasHub: Pretrained Models</a> <a class="nav-link" href="/keras_cv/" role="tab" aria-selected="">KerasCV: Computer Vision Workflows</a> <a class="nav-link" href="/keras_nlp/" role="tab" aria-selected="">KerasNLP: Natural Language Workflows</a> </div> </div> <div class='k-main'> <div class='k-main-top'> <script> function displayDropdownMenu() { e = document.getElementById("nav-menu"); if (e.style.display == "block") { e.style.display = "none"; } else { e.style.display = "block"; document.getElementById("dropdown-nav").style.display = "block"; } } function resetMobileUI() { if (window.innerWidth <= 840) { document.getElementById("nav-menu").style.display = "none"; document.getElementById("dropdown-nav").style.display = "block"; } else { document.getElementById("nav-menu").style.display = "block"; document.getElementById("dropdown-nav").style.display = "none"; } var navmenu = document.getElementById("nav-menu"); var menuheight = navmenu.clientHeight; var kmain = document.getElementById("k-main-id"); kmain.style.minHeight = (menuheight + 100) + 'px'; } window.onresize = resetMobileUI; window.addEventListener("load", (event) => { resetMobileUI() }); </script> <div id='dropdown-nav' onclick="displayDropdownMenu();"> <svg viewBox="-20 -20 120 120" width="60" height="60"> <rect width="100" height="20"></rect> <rect y="30" width="100" height="20"></rect> <rect y="60" width="100" height="20"></rect> </svg> </div> <form class="bd-search d-flex align-items-center k-search-form" id="search-form"> <input type="search" class="k-search-input" id="search-input" placeholder="Search Keras documentation..." aria-label="Search Keras documentation..." autocomplete="off"> <button class="k-search-btn"> <svg width="13" height="13" viewBox="0 0 13 13"><title>search</title><path d="m4.8495 7.8226c0.82666 0 1.5262-0.29146 2.0985-0.87438 0.57232-0.58292 0.86378-1.2877 0.87438-2.1144 0.010599-0.82666-0.28086-1.5262-0.87438-2.0985-0.59352-0.57232-1.293-0.86378-2.0985-0.87438-0.8055-0.010599-1.5103 0.28086-2.1144 0.87438-0.60414 0.59352-0.8956 1.293-0.87438 2.0985 0.021197 0.8055 0.31266 1.5103 0.87438 2.1144 0.56172 0.60414 1.2665 0.8956 2.1144 0.87438zm4.4695 0.2115 3.681 3.6819-1.259 1.284-3.6817-3.7 0.0019784-0.69479-0.090043-0.098846c-0.87973 0.76087-1.92 1.1413-3.1207 1.1413-1.3553 0-2.5025-0.46363-3.4417-1.3909s-1.4088-2.0686-1.4088-3.4239c0-1.3553 0.4696-2.4966 1.4088-3.4239 0.9392-0.92727 2.0864-1.3969 3.4417-1.4088 1.3553-0.011889 2.4906 0.45771 3.406 1.4088 0.9154 0.95107 1.379 2.0924 1.3909 3.4239 0 1.2126-0.38043 2.2588-1.1413 3.1385l0.098834 0.090049z"></path></svg> </button> </form> <script> var form = document.getElementById('search-form'); form.onsubmit = function(e) { e.preventDefault(); var query = document.getElementById('search-input').value; window.location.href = '/search.html?query=' + query; return False } </script> </div> <div class='k-main-inner' id='k-main-id'> <div class='k-location-slug'> ► <a href='/api/'>Keras 3 API documentation</a> / <a href='/api/layers/'>Layers API</a> / Layer weight initializers </div> <div class='k-content'> <h1 id="layer-weight-initializers">Layer weight initializers</h1> <h2 id="usage-of-initializers">Usage of initializers</h2> Initializers define the way to set the initial random weights of Keras layers. The keyword arguments used for passing initializers to layers depends on the layer. Usually, it is simply <code>kernel_initializer</code> and <code>bias_initializer</code>: <div class="codehilite"><pre><code>from keras import layers from keras import initializers layer = layers.Dense( units=64, kernel_initializer=initializers.RandomNormal(stddev=0.01), bias_initializer=initializers.Zeros() ) </code></pre></div> All built-in initializers can also be passed via their string identifier: <div class="codehilite"><pre><code>layer = layers.Dense( units=64, kernel_initializer='random_normal', bias_initializer='zeros' ) </code></pre></div> <hr /> <h2 id="available-initializers">Available initializers</h2> The following built-in initializers are available as part of the <code>keras.initializers</code> module: <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/random_initializers.py#L30">[source]</a> <h3 id="randomnormal-class"><code>RandomNormal</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.RandomNormal(mean=0.0, stddev=0.05, seed=None) </code></pre></div> Random normal initializer. Draws samples from a normal distribution for given parameters. Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = RandomNormal(mean=0.0, stddev=1.0) >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = RandomNormal(mean=0.0, stddev=1.0) >>> layer = Dense(3, kernel_initializer=initializer) </code></pre></div> Arguments <ul> <li>mean: A python scalar or a scalar keras tensor. Mean of the random values to generate.</li> <li>stddev: A python scalar or a scalar keras tensor. Standard deviation of the random values to generate.</li> <li>seed: A Python integer or instance of <code>keras.backend.SeedGenerator</code>. Used to make the behavior of the initializer deterministic. Note that an initializer seeded with an integer or <code>None</code> (unseeded) will produce the same random values across multiple calls. To get different random values across multiple calls, use as seed an instance of <code>keras.backend.SeedGenerator</code>.</li> </ul> <hr /> <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/random_initializers.py#L145">[source]</a> <h3 id="randomuniform-class"><code>RandomUniform</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.RandomUniform(minval=-0.05, maxval=0.05, seed=None) </code></pre></div> Random uniform initializer. Draws samples from a uniform distribution for given parameters. Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = RandomUniform(minval=0.0, maxval=1.0) >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = RandomUniform(minval=0.0, maxval=1.0) >>> layer = Dense(3, kernel_initializer=initializer) </code></pre></div> Arguments <ul> <li>minval: A python scalar or a scalar keras tensor. Lower bound of the range of random values to generate (inclusive).</li> <li>maxval: A python scalar or a scalar keras tensor. Upper bound of the range of random values to generate (exclusive).</li> <li>seed: A Python integer or instance of <code>keras.backend.SeedGenerator</code>. Used to make the behavior of the initializer deterministic. Note that an initializer seeded with an integer or <code>None</code> (unseeded) will produce the same random values across multiple calls. To get different random values across multiple calls, use as seed an instance of <code>keras.backend.SeedGenerator</code>.</li> </ul> <hr /> <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/random_initializers.py#L86">[source]</a> <h3 id="truncatednormal-class"><code>TruncatedNormal</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.TruncatedNormal(mean=0.0, stddev=0.05, seed=None) </code></pre></div> Initializer that generates a truncated normal distribution. The values generated are similar to values from a <code>RandomNormal</code> initializer, except that values more than two standard deviations from the mean are discarded and re-drawn. Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = TruncatedNormal(mean=0., stddev=1.) >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = TruncatedNormal(mean=0., stddev=1.) >>> layer = Dense(3, kernel_initializer=initializer) </code></pre></div> Arguments <ul> <li>mean: A python scalar or a scalar keras tensor. Mean of the random values to generate.</li> <li>stddev: A python scalar or a scalar keras tensor. Standard deviation of the random values to generate.</li> <li>seed: A Python integer or instance of <code>keras.backend.SeedGenerator</code>. Used to make the behavior of the initializer deterministic. Note that an initializer seeded with an integer or <code>None</code> (unseeded) will produce the same random values across multiple calls. To get different random values across multiple calls, use as seed an instance of <code>keras.backend.SeedGenerator</code>.</li> </ul> <hr /> <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/constant_initializers.py#L48">[source]</a> <h3 id="zeros-class"><code>Zeros</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.Zeros() </code></pre></div> Initializer that generates tensors initialized to 0. Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = Zeros() >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = Zeros() >>> layer = Dense(units=3, kernel_initializer=initializer) </code></pre></div> <hr /> <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/constant_initializers.py#L77">[source]</a> <h3 id="ones-class"><code>Ones</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.Ones() </code></pre></div> Initializer that generates tensors initialized to 1. Also available via the shortcut function <code>ones</code>. Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = Ones() >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = Ones() >>> layer = Dense(3, kernel_initializer=initializer) </code></pre></div> <hr /> <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/random_initializers.py#L369">[source]</a> <h3 id="glorotnormal-class"><code>GlorotNormal</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.GlorotNormal(seed=None) </code></pre></div> The Glorot normal initializer, also called Xavier normal initializer. Draws samples from a truncated normal distribution centered on 0 with <code>stddev = sqrt(2 / (fan_in + fan_out))</code> where <code>fan_in</code> is the number of input units in the weight tensor and <code>fan_out</code> is the number of output units in the weight tensor. Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = GlorotNormal() >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = GlorotNormal() >>> layer = Dense(3, kernel_initializer=initializer) </code></pre></div> Arguments <ul> <li>seed: A Python integer or instance of <code>keras.backend.SeedGenerator</code>. Used to make the behavior of the initializer deterministic. Note that an initializer seeded with an integer or <code>None</code> (unseeded) will produce the same random values across multiple calls. To get different random values across multiple calls, use as seed an instance of <code>keras.backend.SeedGenerator</code>.</li> </ul> Reference <ul> <li><a href="http://proceedings.mlr.press/v9/glorot10a.html">Glorot et al., 2010</a></li> </ul> <hr /> <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/random_initializers.py#L320">[source]</a> <h3 id="glorotuniform-class"><code>GlorotUniform</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.GlorotUniform(seed=None) </code></pre></div> The Glorot uniform initializer, also called Xavier uniform initializer. Draws samples from a uniform distribution within <code>[-limit, limit]</code>, where <code>limit = sqrt(6 / (fan_in + fan_out))</code> (<code>fan_in</code> is the number of input units in the weight tensor and <code>fan_out</code> is the number of output units). Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = GlorotUniform() >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = GlorotUniform() >>> layer = Dense(3, kernel_initializer=initializer) </code></pre></div> Arguments <ul> <li>seed: A Python integer or instance of <code>keras.backend.SeedGenerator</code>. Used to make the behavior of the initializer deterministic. Note that an initializer seeded with an integer or <code>None</code> (unseeded) will produce the same random values across multiple calls. To get different random values across multiple calls, use as seed an instance of <code>keras.backend.SeedGenerator</code>.</li> </ul> Reference <ul> <li><a href="http://proceedings.mlr.press/v9/glorot10a.html">Glorot et al., 2010</a></li> </ul> <hr /> <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/random_initializers.py#L524">[source]</a> <h3 id="henormal-class"><code>HeNormal</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.HeNormal(seed=None) </code></pre></div> He normal initializer. It draws samples from a truncated normal distribution centered on 0 with <code>stddev = sqrt(2 / fan_in)</code> where <code>fan_in</code> is the number of input units in the weight tensor. Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = HeNormal() >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = HeNormal() >>> layer = Dense(3, kernel_initializer=initializer) </code></pre></div> Arguments <ul> <li>seed: A Python integer or instance of <code>keras.backend.SeedGenerator</code>. Used to make the behavior of the initializer deterministic. Note that an initializer seeded with an integer or <code>None</code> (unseeded) will produce the same random values across multiple calls. To get different random values across multiple calls, use as seed an instance of <code>keras.backend.SeedGenerator</code>.</li> </ul> Reference <ul> <li><a href="https://arxiv.org/abs/1502.01852">He et al., 2015</a></li> </ul> <hr /> <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/random_initializers.py#L568">[source]</a> <h3 id="heuniform-class"><code>HeUniform</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.HeUniform(seed=None) </code></pre></div> He uniform variance scaling initializer. Draws samples from a uniform distribution within <code>[-limit, limit]</code>, where <code>limit = sqrt(6 / fan_in)</code> (<code>fan_in</code> is the number of input units in the weight tensor). Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = HeUniform() >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = HeUniform() >>> layer = Dense(3, kernel_initializer=initializer) </code></pre></div> Arguments <ul> <li>seed: A Python integer or instance of <code>keras.backend.SeedGenerator</code>. Used to make the behavior of the initializer deterministic. Note that an initializer seeded with an integer or <code>None</code> (unseeded) will produce the same random values across multiple calls. To get different random values across multiple calls, use as seed an instance of <code>keras.backend.SeedGenerator</code>.</li> </ul> Reference <ul> <li><a href="https://arxiv.org/abs/1502.01852">He et al., 2015</a></li> </ul> <hr /> <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/random_initializers.py#L640">[source]</a> <h3 id="orthogonalinitializer-class"><code>OrthogonalInitializer</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.Orthogonal(gain=1.0, seed=None) </code></pre></div> Initializer that generates an orthogonal matrix. If the shape of the tensor to initialize is two-dimensional, it is initialized with an orthogonal matrix obtained from the QR decomposition of a matrix of random numbers drawn from a normal distribution. If the matrix has fewer rows than columns then the output will have orthogonal rows. Otherwise, the output will have orthogonal columns. If the shape of the tensor to initialize is more than two-dimensional, a matrix of shape <code>(shape[0] * ... * shape[n - 2], shape[n - 1])</code> is initialized, where <code>n</code> is the length of the shape vector. The matrix is subsequently reshaped to give a tensor of the desired shape. Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = keras.initializers.Orthogonal() >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = keras.initializers.Orthogonal() >>> layer = keras.layers.Dense(3, kernel_initializer=initializer) </code></pre></div> Arguments <ul> <li>gain: Multiplicative factor to apply to the orthogonal matrix.</li> <li>seed: A Python integer. Used to make the behavior of the initializer deterministic.</li> </ul> Reference <ul> <li><a href="https://openreview.net/forum?id=_wzZwKpTDF_9C">Saxe et al., 2014</a></li> </ul> <hr /> <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/constant_initializers.py#L8">[source]</a> <h3 id="constant-class"><code>Constant</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.Constant(value=0.0) </code></pre></div> Initializer that generates tensors with constant values. Only scalar values are allowed. The constant value provided must be convertible to the dtype requested when calling the initializer. Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = Constant(10.) >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = Constant(10.) >>> layer = Dense(3, kernel_initializer=initializer) </code></pre></div> Arguments <ul> <li>value: A Python scalar.</li> </ul> <hr /> <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/random_initializers.py#L201">[source]</a> <h3 id="variancescaling-class"><code>VarianceScaling</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.VarianceScaling( scale=1.0, mode="fan_in", distribution="truncated_normal", seed=None ) </code></pre></div> Initializer that adapts its scale to the shape of its input tensors. With <code>distribution="truncated_normal" or "untruncated_normal"</code>, samples are drawn from a truncated/untruncated normal distribution with a mean of zero and a standard deviation (after truncation, if used) <code>stddev = sqrt(scale / n)</code>, where <code>n</code> is: <ul> <li>number of input units in the weight tensor, if <code>mode="fan_in"</code></li> <li>number of output units, if <code>mode="fan_out"</code></li> <li>average of the numbers of input and output units, if <code>mode="fan_avg"</code></li> </ul> With <code>distribution="uniform"</code>, samples are drawn from a uniform distribution within <code>[-limit, limit]</code>, where <code>limit = sqrt(3 * scale / n)</code>. Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = VarianceScaling( scale=0.1, mode='fan_in', distribution='uniform') >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = VarianceScaling( scale=0.1, mode='fan_in', distribution='uniform') >>> layer = Dense(3, kernel_initializer=initializer) </code></pre></div> Arguments <ul> <li>scale: Scaling factor (positive float).</li> <li>mode: One of <code>"fan_in"</code>, <code>"fan_out"</code>, <code>"fan_avg"</code>.</li> <li>distribution: Random distribution to use. One of <code>"truncated_normal"</code>, <code>"untruncated_normal"</code>, or <code>"uniform"</code>.</li> <li>seed: A Python integer or instance of <code>keras.backend.SeedGenerator</code>. Used to make the behavior of the initializer deterministic. Note that an initializer seeded with an integer or <code>None</code> (unseeded) will produce the same random values across multiple calls. To get different random values across multiple calls, use as seed an instance of <code>keras.backend.SeedGenerator</code>.</li> </ul> <hr /> <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/random_initializers.py#L422">[source]</a> <h3 id="lecunnormal-class"><code>LecunNormal</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.LecunNormal(seed=None) </code></pre></div> Lecun normal initializer. Initializers allow you to pre-specify an initialization strategy, encoded in the Initializer object, without knowing the shape and dtype of the variable being initialized. Draws samples from a truncated normal distribution centered on 0 with <code>stddev = sqrt(1 / fan_in)</code> where <code>fan_in</code> is the number of input units in the weight tensor. Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = LecunNormal() >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = LecunNormal() >>> layer = Dense(3, kernel_initializer=initializer) </code></pre></div> Arguments <ul> <li>seed: A Python integer or instance of <code>keras.backend.SeedGenerator</code>. Used to make the behavior of the initializer deterministic. Note that an initializer seeded with an integer or <code>None</code> (unseeded) will produce the same random values across multiple calls. To get different random values across multiple calls, use as seed an instance of <code>keras.backend.SeedGenerator</code>.</li> </ul> Reference <ul> <li><a href="https://arxiv.org/abs/1706.02515">Klambauer et al., 2017</a></li> </ul> <hr /> <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/random_initializers.py#L475">[source]</a> <h3 id="lecununiform-class"><code>LecunUniform</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.LecunUniform(seed=None) </code></pre></div> Lecun uniform initializer. Draws samples from a uniform distribution within <code>[-limit, limit]</code>, where <code>limit = sqrt(3 / fan_in)</code> (<code>fan_in</code> is the number of input units in the weight tensor). Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = LecunUniform() >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = LecunUniform() >>> layer = Dense(3, kernel_initializer=initializer) </code></pre></div> Arguments <ul> <li>seed: A Python integer or instance of <code>keras.backend.SeedGenerator</code>. Used to make the behavior of the initializer deterministic. Note that an initializer seeded with an integer or <code>None</code> (unseeded) will produce the same random values across multiple calls. To get different random values across multiple calls, use as seed an instance of <code>keras.backend.SeedGenerator</code>.</li> </ul> Reference <ul> <li><a href="https://arxiv.org/abs/1706.02515">Klambauer et al., 2017</a></li> </ul> <hr /> <a href="https://github.com/keras-team/keras/tree/v3.6.0/keras/src/initializers/constant_initializers.py#L108">[source]</a> <h3 id="identity-class"><code>Identity</code> class</h3> <div class="codehilite"><pre><code>keras.initializers.IdentityInitializer(gain=1.0) </code></pre></div> Initializer that generates the identity matrix. Only usable for generating 2D matrices. Examples <div class="codehilite"><pre><code>>>> # Standalone usage: >>> initializer = Identity() >>> values = initializer(shape=(2, 2)) </code></pre></div> <div class="codehilite"><pre><code>>>> # Usage in a Keras layer: >>> initializer = Identity() >>> layer = Dense(3, kernel_initializer=initializer) </code></pre></div> Arguments <ul> <li>gain: Multiplicative factor to apply to the identity matrix.</li> </ul> <hr /> <h2 id="creating-custom-initializers">Creating custom initializers</h2> <h3 id="simple-callables">Simple callables</h3> You can pass a custom callable as initializer. It must take the arguments <code>shape</code> (shape of the variable to initialize) and <code>dtype</code> (dtype of generated values): <div class="codehilite"><pre><code>def my_init(shape, dtype=None): return keras.random.normal(shape, dtype=dtype) layer = Dense(64, kernel_initializer=my_init) </code></pre></div> <h3 id="initializer-subclasses"><code>Initializer</code> subclasses</h3> If you need to configure your initializer via various arguments (e.g. <code>stddev</code> argument in <code>RandomNormal</code>), you should implement it as a subclass of <code>keras.initializers.Initializer</code>. Initializers should implement a <code>__call__</code> method with the following signature: <div class="codehilite"><pre><code>def __call__(self, shape, dtype=None)`: # returns a tensor of shape `shape` and dtype `dtype` # containing values drawn from a distribution of your choice. </code></pre></div> Optionally, you an also implement the method <code>get_config</code> and the class method <code>from_config</code> in order to support serialization – just like with any Keras object. Here's a simple example: a random normal initializer. <div class="codehilite"><pre><code>class ExampleRandomNormal(keras.initializers.Initializer): def __init__(self, mean, stddev): self.mean = mean self.stddev = stddev def __call__(self, shape, dtype=None)`: return keras.random.normal( shape, mean=self.mean, stddev=self.stddev, dtype=dtype) def get_config(self): # To support serialization return {'mean': self.mean, 'stddev': self.stddev} </code></pre></div> Note that we don't have to implement <code>from_config</code> in the example above since the constructor arguments of the class the keys in the config returned by <code>get_config</code> are the same. In this case, the default <code>from_config</code> works fine. </div> <div class='k-outline'> <div class='k-outline-depth-1'> <a href='#layer-weight-initializers'>Layer weight initializers</a> </div> <div class='k-outline-depth-2'> ◆ <a href='#usage-of-initializers'>Usage of initializers</a> </div> <div class='k-outline-depth-2'> ◆ <a href='#available-initializers'>Available initializers</a> </div> <div class='k-outline-depth-3'> <a href='#randomnormal-class'><code>RandomNormal</code> class</a> </div> <div class='k-outline-depth-3'> <a href='#randomuniform-class'><code>RandomUniform</code> class</a> </div> <div class='k-outline-depth-3'> <a href='#truncatednormal-class'><code>TruncatedNormal</code> class</a> </div> <div class='k-outline-depth-3'> <a href='#zeros-class'><code>Zeros</code> class</a> </div> <div class='k-outline-depth-3'> <a href='#ones-class'><code>Ones</code> class</a> </div> <div class='k-outline-depth-3'> <a href='#glorotnormal-class'><code>GlorotNormal</code> class</a> </div> <div class='k-outline-depth-3'> <a href='#glorotuniform-class'><code>GlorotUniform</code> class</a> </div> <div class='k-outline-depth-3'> <a href='#henormal-class'><code>HeNormal</code> class</a> </div> <div class='k-outline-depth-3'> <a href='#heuniform-class'><code>HeUniform</code> class</a> </div> <div class='k-outline-depth-3'> <a href='#orthogonalinitializer-class'><code>OrthogonalInitializer</code> class</a> </div> <div class='k-outline-depth-3'> <a href='#constant-class'><code>Constant</code> class</a> </div> <div class='k-outline-depth-3'> <a href='#variancescaling-class'><code>VarianceScaling</code> class</a> </div> <div class='k-outline-depth-3'> <a href='#lecunnormal-class'><code>LecunNormal</code> class</a> </div> <div class='k-outline-depth-3'> <a href='#lecununiform-class'><code>LecunUniform</code> class</a> </div> <div class='k-outline-depth-3'> <a href='#identity-class'><code>Identity</code> class</a> </div> <div class='k-outline-depth-2'> ◆ <a href='#creating-custom-initializers'>Creating custom initializers</a> </div> <div class='k-outline-depth-3'> <a href='#simple-callables'>Simple callables</a> </div> <div class='k-outline-depth-3'> <a href='#initializer-subclasses'><code>Initializer</code> subclasses</a> </div> </div> </div> </div> </div> </body> <footer style="float: left; width: 100%; padding: 1em; border-top: solid 1px #bbb;"> <a href="https://policies.google.com/terms">Terms</a> | <a href="https://policies.google.com/privacy">Privacy</a> </footer> </html>