{
    "info": {
        "author": "Jonathan Raiman",
        "author_email": "jraiman at mit dot edu",
        "bugtrack_url": null,
        "classifiers": [
            "Intended Audience :: Science/Research",
            "Operating System :: OS Independent",
            "Programming Language :: Python :: 3.3",
            "Topic :: Scientific/Engineering :: Artificial Intelligence",
            "Topic :: Scientific/Engineering :: Mathematics",
            "Topic :: Text Processing :: Linguistic"
        ],
        "description": "# Gradient Optimizers\n\nOptimize you Theano Models with [Adagrad](http://www.magicbroom.info/Papers/DuchiHaSi10.pdf), Hessian Free optimization, or linear updates.\n\n\n    pip3 install gradient-optimizers\n\n\nSee example notebook (TBD) for tutorial.\n\nTwo classes **GradientModel**, and **GradientHFModel**, for optimizing gradient\nbased models (specifically built with indexed parameters in mind (e.g.\nfor language models))\n\n## GradientModel\n\nA gradient model for updating your model with\nhessian free, adagrad, or linear decay updates.\n\nYou will need to define the following attributes,\nand fill them as appropriate:\n    \n    # a forward method for getting errors:\n    projection = self.projection_function(ivector <indices/>)\n\n    # a cost function (that takes the result of projection function and labels as input)\n    # and returns a symbolic differentiable theano variable\n    self.cost_function(projection, ivector <label/>).sum()\n\n    self.params = []\n    self.indexed_params = set()\n\n    self._l2_regularization = True / False\n\n    self.store_max_updates = True / False\n\n    # set this theano setting\n    self.theano_mode = \"FAST_RUN\"\n\n    # set this theano setting\n    self.disconnected_inputs = 'ignore' / None\n\n    # if L2 is true store this parameter:\n    self._l2_regularization_parameter = theano.shared(np.float64(l2_regularization).astype(REAL), name='l2_regularization_parameter')\n\nUpon initialization you must run:\n\n    self._select_update_mechanism(update_method_name)\n\n    # then to compile this mechanism:\n    self.create_update_fun()\n\nThe update methods expect the input to be of the form:\n\n    ivector <indices/>, ivector <labels/>\n\nIf this is not the case you can modify them as appropriate.\n\n## GradientHFModel\n\nImplements an symbolic one step of hessian-free [1]\noptimization that approximates the curvature,\nrequires a _compute_cost method that takes an example\nas input or a _compute_cost_gradients that returns\ngradients for each example provided.\n\nModel should have a params property containing symbolic\ntheano variables.\n\n[[1] James Martens, ``Deep learning via Hessian-free optimization\", ICML 2010](http://www.icml2010.org/papers/458.pdf)\n\nMake sure the following parameters are not tampered with:\n\n    self._additional_params\n\n    self._num_updates",
        "description_content_type": null,
        "docs_url": null,
        "download_url": "https://github.com/JonathanRaiman/gradient_optimizers",
        "downloads": {
            "last_day": -1,
            "last_month": -1,
            "last_week": -1
        },
        "home_page": "https://github.com/JonathanRaiman/gradient_optimizers",
        "keywords": "Machine Learning,Gradient Descent,NLP,Optimization,Hessian Free optimization",
        "license": "MIT",
        "maintainer": null,
        "maintainer_email": null,
        "name": "gradient-optimizers",
        "package_url": "https://pypi.org/project/gradient-optimizers/",
        "platform": "any",
        "project_url": "https://pypi.org/project/gradient-optimizers/",
        "project_urls": {
            "Download": "https://github.com/JonathanRaiman/gradient_optimizers",
            "Homepage": "https://github.com/JonathanRaiman/gradient_optimizers"
        },
        "release_url": "https://pypi.org/project/gradient-optimizers/0.0.4/",
        "requires_dist": null,
        "requires_python": null,
        "summary": "Python package for wrapping gradient optimizers for models in Theano",
        "version": "0.0.4"
    },
    "last_serial": 1289311,
    "releases": {
        "0.0.1": [
            {
                "comment_text": "",
                "digests": {
                    "md5": "a26fb25aa269cd67d366849a18c726da",
                    "sha256": "4357aeacb7c7a067c83764161697058eae598bb77bf1189dc48dd6e1afc22e6d"
                },
                "downloads": -1,
                "filename": "gradient-optimizers-0.0.1.tar.gz",
                "has_sig": false,
                "md5_digest": "a26fb25aa269cd67d366849a18c726da",
                "packagetype": "sdist",
                "python_version": "source",
                "requires_python": null,
                "size": 8602,
                "upload_time": "2014-10-30T22:34:04",
                "url": "https://files.pythonhosted.org/packages/3e/f6/204264dba9f61f13e9f5412cafef1f2f05539af476b3ec13b9c43aeb807e/gradient-optimizers-0.0.1.tar.gz"
            }
        ],
        "0.0.2": [
            {
                "comment_text": "",
                "digests": {
                    "md5": "bf4e69456da104c77f5e8e420e81237f",
                    "sha256": "0c0d7b53f61dff6168817535f967ec602237176ec09d12d2d18bf9389ad6854a"
                },
                "downloads": -1,
                "filename": "gradient-optimizers-0.0.2.tar.gz",
                "has_sig": false,
                "md5_digest": "bf4e69456da104c77f5e8e420e81237f",
                "packagetype": "sdist",
                "python_version": "source",
                "requires_python": null,
                "size": 8750,
                "upload_time": "2014-10-30T22:46:00",
                "url": "https://files.pythonhosted.org/packages/b3/15/739c19a273c96b9d0b3773083533c4e6bc7304b82a676fbe58154cef072d/gradient-optimizers-0.0.2.tar.gz"
            }
        ],
        "0.0.3": [
            {
                "comment_text": "",
                "digests": {
                    "md5": "3a2618f0479842bbaacf3c4e646b5ef0",
                    "sha256": "fb73a3265416a53199c27bb12fa606f13c604da142e94beddf61128e6813de06"
                },
                "downloads": -1,
                "filename": "gradient-optimizers-0.0.3.tar.gz",
                "has_sig": false,
                "md5_digest": "3a2618f0479842bbaacf3c4e646b5ef0",
                "packagetype": "sdist",
                "python_version": "source",
                "requires_python": null,
                "size": 9143,
                "upload_time": "2014-10-30T22:55:49",
                "url": "https://files.pythonhosted.org/packages/b1/c3/d7e02b3efa96a4685e01c2c79e81507b6cee688d16495e4fe1473101f491/gradient-optimizers-0.0.3.tar.gz"
            }
        ],
        "0.0.4": [
            {
                "comment_text": "",
                "digests": {
                    "md5": "2f09e8041063a3dfd53dd364b2418844",
                    "sha256": "9ccfeb3b13297ade4506c1cdb1574d1f09f63018df7046fb5c0f9450a3bf9d4f"
                },
                "downloads": -1,
                "filename": "gradient-optimizers-0.0.4.tar.gz",
                "has_sig": false,
                "md5_digest": "2f09e8041063a3dfd53dd364b2418844",
                "packagetype": "sdist",
                "python_version": "source",
                "requires_python": null,
                "size": 9186,
                "upload_time": "2014-10-30T23:07:07",
                "url": "https://files.pythonhosted.org/packages/04/72/2a6824a6af69c4fa8eb2aa2c3348625490bb58d5b18f149b28aa7a719643/gradient-optimizers-0.0.4.tar.gz"
            }
        ]
    },
    "urls": [
        {
            "comment_text": "",
            "digests": {
                "md5": "2f09e8041063a3dfd53dd364b2418844",
                "sha256": "9ccfeb3b13297ade4506c1cdb1574d1f09f63018df7046fb5c0f9450a3bf9d4f"
            },
            "downloads": -1,
            "filename": "gradient-optimizers-0.0.4.tar.gz",
            "has_sig": false,
            "md5_digest": "2f09e8041063a3dfd53dd364b2418844",
            "packagetype": "sdist",
            "python_version": "source",
            "requires_python": null,
            "size": 9186,
            "upload_time": "2014-10-30T23:07:07",
            "url": "https://files.pythonhosted.org/packages/04/72/2a6824a6af69c4fa8eb2aa2c3348625490bb58d5b18f149b28aa7a719643/gradient-optimizers-0.0.4.tar.gz"
        }
    ]
}