{
    "info": {
        "author": "Lin Xiao",
        "author_email": "lin.xiao@gmail.com",
        "bugtrack_url": null,
        "classifiers": [],
        "description": "# Accelerated Bregman Proximal Gradient Methods\n\nAccelerated first-order algorithms for solving relatively-smooth convex optimization problems of the form\n\n    minimize { f(x) + P(x) | x in C }\n\nwith a reference function h(x), where\n\n* h(x) is convex and essentially smooth on C\n* f(x) is convex and differentiable, and L-smooth relative to h(x), that is, f(x)-L*h(x) is convex\n* P(x) is convex and closed (lower semi-continuous)\n* C is a closed convex set\n\n### Implemented algorithms in [HRX2018](https://arxiv.org/abs/1808.03045)\n\n* BPG_LS (Bregman proximal gradient) method with line search\n* ABPG (Accelerated BPG) method\n* ABPG-expo (ABPG with exponent adaption)\n* ABPG-gain (ABPG with gain adaption)\n* ABDA (Accelerated Bregman dual averaging) method\n\n## Installation\n\nClone or fork from GitHub. Or install from PyPI:\n\n    pip install accbpg\n\n## Usage\n\n    import accbpg\n\n    # generate a random instance of D-optimal design problem\n    f, h, L, x0 = accbpg.D_opt_design(80, 200)\n\n    # solve the problem instance using BPG with line search\n    x1, F1, G1 = accbpg.BPG_LS(f, h, L, x0, maxitrs=1000, verskip=100)\n\n    # solve it again using ABPG_gain with gamma=2\n    x2, F2, G2, D2 = accbpg.ABPG_gain(f, h, L, 2, x0, maxitrs=1000, verbskip=100)\n\ncompare the two methods by visualization\n\n    import matplotlib.pyplot as plt\n    Fmin = min(F1.min(), F2.min())\n    plt.semilogy(range(len(F1)), F1-Fmin, range(len(F2)), F2-Fmin)\n\n## Examples in [HRX2018](https://arxiv.org/abs/1808.03045)\n\n**D-optimal experiment design**\n    \n    import accbpg.ex_D_opt\n\n**Nonnegative regression with KL-divergence**\n\n    import accbpg.ex_KL_regr\n\n**Poisson linear inverse problems**\n\n    import accbpg.ex_PoissonL1\n    import accbpg.ex_PoissonL2",
        "description_content_type": "",
        "docs_url": null,
        "download_url": "",
        "downloads": {
            "last_day": -1,
            "last_month": -1,
            "last_week": -1
        },
        "home_page": "https://github.com/Microsoft/accbpg",
        "keywords": "",
        "license": "MIT",
        "maintainer": "",
        "maintainer_email": "",
        "name": "accbpg",
        "package_url": "https://pypi.org/project/accbpg/",
        "platform": "",
        "project_url": "https://pypi.org/project/accbpg/",
        "project_urls": {
            "Homepage": "https://github.com/Microsoft/accbpg"
        },
        "release_url": "https://pypi.org/project/accbpg/0.1/",
        "requires_dist": null,
        "requires_python": "",
        "summary": "Accelerated Bregman proximal gradient (ABPG) methods",
        "version": "0.1"
    },
    "last_serial": 4162144,
    "releases": {
        "0.1": [
            {
                "comment_text": "",
                "digests": {
                    "md5": "f87201ec84c29d2d6e4a8f7df5f0253d",
                    "sha256": "d3f5bad00d00b142a05eb4dfb3cebf9fb43b0bc14d6d425131f0189c86b49045"
                },
                "downloads": -1,
                "filename": "accbpg-0.1.tar.gz",
                "has_sig": false,
                "md5_digest": "f87201ec84c29d2d6e4a8f7df5f0253d",
                "packagetype": "sdist",
                "python_version": "source",
                "requires_python": null,
                "size": 11231,
                "upload_time": "2018-08-12T17:56:43",
                "url": "https://files.pythonhosted.org/packages/22/87/f20896f715a24d5e6b565d2bae30b30926d7541c9b607d0ff920393baa2e/accbpg-0.1.tar.gz"
            }
        ]
    },
    "urls": [
        {
            "comment_text": "",
            "digests": {
                "md5": "f87201ec84c29d2d6e4a8f7df5f0253d",
                "sha256": "d3f5bad00d00b142a05eb4dfb3cebf9fb43b0bc14d6d425131f0189c86b49045"
            },
            "downloads": -1,
            "filename": "accbpg-0.1.tar.gz",
            "has_sig": false,
            "md5_digest": "f87201ec84c29d2d6e4a8f7df5f0253d",
            "packagetype": "sdist",
            "python_version": "source",
            "requires_python": null,
            "size": 11231,
            "upload_time": "2018-08-12T17:56:43",
            "url": "https://files.pythonhosted.org/packages/22/87/f20896f715a24d5e6b565d2bae30b30926d7541c9b607d0ff920393baa2e/accbpg-0.1.tar.gz"
        }
    ]
}