{
    "info": {
        "author": "Isaac Kriegman",
        "author_email": "zackriegman@gmail.com",
        "bugtrack_url": null,
        "classifiers": [
            "Development Status :: 2 - Pre-Alpha",
            "Environment :: Console",
            "Intended Audience :: Developers",
            "Intended Audience :: Education",
            "Intended Audience :: End Users/Desktop",
            "Intended Audience :: Science/Research",
            "License :: OSI Approved :: MIT License",
            "Programming Language :: Python :: 2.7",
            "Programming Language :: Python :: 3.4",
            "Topic :: Scientific/Engineering :: Artificial Intelligence",
            "Topic :: Scientific/Engineering :: Image Recognition",
            "Topic :: Scientific/Engineering :: Information Analysis",
            "Topic :: Software Development :: Libraries"
        ],
        "description": "********************************************\npydnn: deep neural network library in Python\n********************************************\n\npydnn is a deep neural network library written in Python using `Theano <http://deeplearning.net/software/theano/>`_ (symbolic math and optimizing compiler package).  I wrote it as a learning project while competing in `Kaggle's National Data Science Bowl <http://www.datasciencebowl.com/>`_ in March 2015 (where it produced an entry finishing in the `top 6% <http://www.kaggle.com/c/datasciencebowl/leaderboard/private>`_) and plan to continue developing it by adding support for the most important deep learning techniques (including RNNs).\n\n============\nDesign Goals\n============\n\n* **Simplicity**\n    Wherever possible simplify code to make it a clear expression of underlying deep learning algorithms.  Minimize cognitive overhead, so that it is easy for someone who has completed the `deeplearning.net tutorials <http://deeplearning.net/tutorial/>`_ to pickup this library as a next step and easily start learning about, using, and coding more advanced techniques.\n\n* **Completeness**\n    Include all the important and popular techniques for effective deep learning and **not** techniques with more marginal or ambiguous benefit.\n\n* **Ease of use**\n    Make preparing a dataset, building a model and training a deep network only a few lines of code; enable users to work with NumPy rather than Theano.\n\n* **Performance**\n    Should be roughly on par with other Theano neural net libraries so that pydnn is a viable choice for computationally intensive deep learning.\n\n========\nFeatures\n========\n\n* High performance GPU training (courtesy of Theano)\n* Quick start tools to instantly get started training on `inexpensive <http://aws.amazon.com/ec2/pricing/>`_ Amazon EC2 GPU instances.\n* Implementations of important new techniques recently reported in the literature:\n    * `Batch Normalization <http://arxiv.org/pdf/1502.03167v3.pdf>`_\n    * `Parametric ReLU <http://arxiv.org/pdf/1502.01852.pdf>`_ activation function,\n    * `Adam <http://arxiv.org/pdf/1412.6980v4.pdf>`_ optimization\n    * `AdaDelta <http://arxiv.org/pdf/1212.5701v1.pdf>`_ optimization\n    * etc.\n* Implementations of standard deep learning techniques:\n    * Stochastic Gradient Descent with Momentum\n    * Dropout\n    * convolutions with max-pooling using overlapping windows\n    * ReLU/Tanh/sigmoid activation functions\n    * etc.\n\n=====\nUsage\n=====\n\nFirst download and unzip raw image data from somewhere (e.g. Kaggle). Then::\n\n    import pydnn\n    import numpy as np\n    rng = np.random.RandomState(e.rng_seed)\n\n    # build data, split into training/validation sets, preprocess\n    train_dir = 'home\\ubuntu\\train'\n    data = pydnn.data.DirectoryLabeledImageSet(train_dir).build()\n    data = pydnn.preprocess.split_training_data(data, 64, 80, 15, 5)\n    resizer = pydnn.preprocess.StretchResizer()\n    pre = pydnn.preprocess.Rotator360(data, (64, 64), resizer, rng)\n\n    # build the neural network\n    net = pydnn.nn.NN(pre, 'images', 121, 64, rng, pydnn.nn.relu)\n    net.add_convolution(72, (7, 7), (2, 2))\n    net.add_dropout()\n    net.add_convolution(128, (5, 5), (2, 2))\n    net.add_dropout()\n    net.add_convolution(128, (3, 3), (2, 2))\n    net.add_dropout()\n    net.add_hidden(3072)\n    net.add_dropout()\n    net.add_hidden(3072)\n    net.add_dropout()\n    net.add_logistic()\n\n    # train the network\n    lr = pydnn.nn.Adam(learning_rate=pydnn.nn.LearningRateDecay(\n                learning_rate=0.006,\n                decay=.1))\n    net.train(lr)\n\nFrom raw data to trained network (including specifying\nnetwork architecture) in 25 lines of code.\n\n\n================\nShort Term Goals\n================\n\n* Implement popular RNN techniques.\n* Integrate with Amazon EC2 clustering software (such as `StarCluster <http://star.mit.edu/cluster/>`_).\n* Integrate with hyper-parameter optimization frameworks (such as `Spearmint <https://github.com/JasperSnoek/spearmint>`_ and `hyperopt <https://github.com/hyperopt/hyperopt>`_).\n\n=======\nAuthors\n=======\n\nIsaac Kriegman",
        "description_content_type": null,
        "docs_url": null,
        "download_url": "UNKNOWN",
        "downloads": {
            "last_day": -1,
            "last_month": -1,
            "last_week": -1
        },
        "home_page": "https://github.com/zackriegman/pydnn",
        "keywords": "neural network deep learning AI machine learning",
        "license": "MIT",
        "maintainer": null,
        "maintainer_email": null,
        "name": "pydnn",
        "package_url": "https://pypi.org/project/pydnn/",
        "platform": "UNKNOWN",
        "project_url": "https://pypi.org/project/pydnn/",
        "project_urls": {
            "Download": "UNKNOWN",
            "Homepage": "https://github.com/zackriegman/pydnn"
        },
        "release_url": "https://pypi.org/project/pydnn/0.0.dev/",
        "requires_dist": null,
        "requires_python": null,
        "summary": "deep neural network library in Python",
        "version": "0.0.dev"
    },
    "last_serial": 1478176,
    "releases": {
        "0.0.dev": [
            {
                "comment_text": "",
                "digests": {
                    "md5": "6fbed30f08d8c0a780a3e3fb60de7a95",
                    "sha256": "bf5b7a799fe8d6124b15bb18ea5cf5182a05d411db28b1aa21bd16369f5de113"
                },
                "downloads": -1,
                "filename": "pydnn-0.0.dev-py2.py3-none-any.whl",
                "has_sig": false,
                "md5_digest": "6fbed30f08d8c0a780a3e3fb60de7a95",
                "packagetype": "bdist_wheel",
                "python_version": "py2.py3",
                "requires_python": null,
                "size": 75067,
                "upload_time": "2015-03-26T04:42:20",
                "url": "https://files.pythonhosted.org/packages/ab/8b/fde2a005c42035569835ae962a49e958f75e41000bed86ee89f10536fa67/pydnn-0.0.dev-py2.py3-none-any.whl"
            }
        ]
    },
    "urls": [
        {
            "comment_text": "",
            "digests": {
                "md5": "6fbed30f08d8c0a780a3e3fb60de7a95",
                "sha256": "bf5b7a799fe8d6124b15bb18ea5cf5182a05d411db28b1aa21bd16369f5de113"
            },
            "downloads": -1,
            "filename": "pydnn-0.0.dev-py2.py3-none-any.whl",
            "has_sig": false,
            "md5_digest": "6fbed30f08d8c0a780a3e3fb60de7a95",
            "packagetype": "bdist_wheel",
            "python_version": "py2.py3",
            "requires_python": null,
            "size": 75067,
            "upload_time": "2015-03-26T04:42:20",
            "url": "https://files.pythonhosted.org/packages/ab/8b/fde2a005c42035569835ae962a49e958f75e41000bed86ee89f10536fa67/pydnn-0.0.dev-py2.py3-none-any.whl"
        }
    ]
}