コード例 #1
0
def test_solve_nestedcs_bce():
    """test nested cs can be solved
    """
    import numpy as np
    from crpm.setup_nestedcs import setup_nestedcs
    from crpm.fwdprop import fwdprop
    from crpm.lossfunctions import loss
    from crpm.gradientdecent import gradientdecent

    #init numpy seed
    np.random.seed(40017)

    #setup model
    model, data = setup_nestedcs()

    #calculate initial binary cross entropy error
    pred, _ = fwdprop(data[0:2, ], model)
    icost, _ = loss("bce", pred, data[-1, ])

    #train model
    pred, cost, _ = gradientdecent(model, data[0:2, ], data[-1, ], "bce")

    #print(model)
    #print(icost)
    #print(cost)
    assert icost > cost
    assert cost < .29
コード例 #2
0
def will_roc_will_plot():
    """ test if roc output will plot properly"""
    import os
    import matplotlib.pyplot as plt
    from crpm.setup_nestedcs import setup_nestedcs
    from crpm.gradientdecent import gradientdecent
    from crpm.analyzebinaryclassifier import analyzebinaryclassifier

    #setup model
    model, data = setup_nestedcs()

    #train model
    pred, _, _ = gradientdecent(model, data[0:2, ], data[-1, ], "mse")

    #analyze binary classifier
    roc, _ = analyzebinaryclassifier(pred, data[-1, ])

    plt.scatter(*zip(*roc))

    #remove any previously saved files if they exist
    if os.path.exists("nestedcs_roc.png"):
        os.remove("nestedcs_roc.png")

    #save file
    plt.savefig("nestedcs_roc.png")
    #plt.show()

    #assert file was created then remove
    assert os.path.exists("nestedcs_roc.png")
    os.remove("nestedcs_roc.png")
コード例 #3
0
def test_solve_nestedcs_bce():
    """test nested cs can be solved
    """
    import numpy as np
    from crpm.setup_nestedcs import setup_nestedcs
    from crpm.fwdprop import fwdprop
    from crpm.lossfunctions import loss
    from crpm.langevindynamics import langevindynamics

    #init numpy seed
    np.random.seed(40017)

    #setup model
    model, data = setup_nestedcs()

    #calculate initial binary cross entropy error
    pred, __ = fwdprop(data[0:2,], model)
    icost, __ = loss("bce", pred, data[-1,])

    #train model
    pred, cost = langevindynamics(model, data[0:2,], data[-1,], "bce",
                                  maxepoch=int(2E3), maxbuffer=int(1E2))

    #print(model)
    #print(icost)
    #print(cost)
    assert icost > cost
    assert cost < .29
コード例 #4
0
ファイル: test_som.py プロジェクト: dmontemayor/CRPM
def r_test_som2d_init_pca():
    """test inital node coordinates for 2D SOM follow known PCA for nested Cs data.

    We know for Nested Cs data - first 2 PCs point roughly in x and y directions
    """

    import numpy as np
    from crpm.setup_nestedcs import setup_nestedcs
    from crpm.fwdprop import fwdprop
    from crpm.som import init_som
    from crpm.som import som

    #init numpy seed
    np.random.seed(40017)

    #setup model with nested Cs
    model, data = setup_nestedcs()

    #update state of model
    _, state = fwdprop(data[0:2, ], model)

    #create and init 2D map with model and its current state
    map, _ = init_som(model, state, n=100, nx=10, ny=10, hcp=True)

    #check nodes map to points along x and y direction with reasonable span
    xmin = min(data[0, ])
    xmax = max(data[0, ])
    xvar = np.var(data[0, ])
    wxmin = min(map[-1]["weight"][:, 0])
    wxmax = max(map[-1]["weight"][:, 0])
    wxdel = map[-1]["weight"][-1, 0] - map[-1]["weight"][0, 0]
    ymin = min(data[1, ])
    ymax = max(data[1, ])
    yvar = np.var(data[1, ])
    wymin = min(map[-1]["weight"][:, 1])
    wymax = max(map[-1]["weight"][:, 1])
    wydel = map[-1]["weight"][-1, 1] - map[-1]["weight"][0, 1]
    slope = (ymax - ymin) / (xmax - xmin)

    #assert nodes point in the correct x direction
    assert (np.sign(wxmax - wxmin) == np.sign(xmax - xmin))
    #assert nodes point in the correct y direction
    assert (np.sign(wymax - wymin) == np.sign(ymax - ymin))
    #assert fist and last nodes point with slope similar to data range
    assert (abs(wydel / wxdel - slope) / slope < 0.4)
    #asset span of x points is comperable to span of input x points
    assert (abs(np.var(map[-1]["weight"][:, 0]) - xvar) / xvar < 1.0)
    #asset span of y points is comperable to span of input y points
    assert (abs(np.var(map[-1]["weight"][:, 1]) - yvar) / yvar < 1.0)
コード例 #5
0
def test_encode_nestedcs():
    """test nested cs data can be encoded
    """

    import numpy as np
    from crpm.setup_nestedcs import setup_nestedcs
    from crpm.contrastivedivergence import contrastivedivergence
    from crpm.dynamics import computecost
    from crpm.analyzebinaryclassifier import analyzebinaryclassifier
    from crpm.ffn_bodyplan import stack_new_layer
    from crpm.gradientdecent import gradientdecent
    #init numpy seed
    np.random.seed(40017)

    #setup model
    model, data = setup_nestedcs()

    #remove discriminating layer
    prototype = model[0:-1]

    #explicitly remove labels from data
    #labels = data[2, :]
    data = data[0:2, :]

    #return untrained autoencoder
    _, autoencoder = contrastivedivergence(prototype, data, maxepoch=0)

    #calculate initial mean squared error
    pred, icost = computecost(autoencoder, data, data, "mse")

    #train model
    _, autoencoder = contrastivedivergence(prototype,
                                           data,
                                           ncd=10,
                                           nadj=10,
                                           maxepoch=1000,
                                           momentum=0.1,
                                           batchsize=10,
                                           finetune=6)

    #calculate final mean squared error
    pred, cost = computecost(autoencoder, data, data, "mse")

    #assert learning is taking place
    assert icost > cost
コード例 #6
0
def test_solved_nestedcs_bce():
    """test solved nestedcs will produce optimal threashold
    """
    from crpm.setup_nestedcs import setup_nestedcs
    from crpm.gradientdecent import gradientdecent
    from crpm.analyzebinaryclassifier import analyzebinaryclassifier

    #setup model
    model, data = setup_nestedcs()

    #train model
    pred, _, _ = gradientdecent(model, data[0:2, ], data[-1, ], "bce")

    #analyze binary classifier
    _, report = analyzebinaryclassifier(pred, data[-1, ])

    print(report)
    assert report["AreaUnderCurve"] >= .95
    assert report["Accuracy"] >= .87
    assert report["Accuracy"] < 1.0
    assert report["F1Score"] >= .86
コード例 #7
0
def test_solved_nestedcs():
    """test solved nestedcs will produce optimal threashold
    """
    import numpy as np
    from crpm.setup_nestedcs import setup_nestedcs
    from crpm.gradientdecent import gradientdecent
    from crpm.analyzebinaryclassifier import analyzebinaryclassifier

    #init numpy seed
    np.random.seed(40017)

    #setup model
    model, data = setup_nestedcs()

    #train model
    pred, _, _ = gradientdecent(model, data[0:2, ], data[-1, ], "mse")

    #analyze binary classifier
    _, report = analyzebinaryclassifier(pred, data[-1, ])

    assert report["AreaUnderCurve"] >= .92
    assert report["Accuracy"] >= .80
    assert report["F1Score"] >= .80
コード例 #8
0
ファイル: test_som.py プロジェクト: dmontemayor/CRPM
def test_som1d_init_pca():
    """test inital node coordinates for 1D SOM follow known PCA for nested Cs data.

    We know for Nested Cs data - first 2 PCs point roughly in x and y directions
    """

    import numpy as np
    from crpm.setup_nestedcs import setup_nestedcs
    from crpm.fwdprop import fwdprop
    from crpm.som import init_som
    from crpm.som import som

    #init numpy seed
    np.random.seed(40017)

    #setup model with nested Cs
    model, data = setup_nestedcs()

    #update state of model
    _, state = fwdprop(data[0:2, ], model)

    #create and init 1D map with model and its current state
    map, _ = init_som(model, state)

    #check nodes map to points along x direction with reasonable span
    xmin = min(data[0, ])
    xmax = max(data[0, ])
    wmin = min(map[-1]["weight"][:, 0])
    wmax = max(map[-1]["weight"][:, 0])
    xvar = np.var(data[0, ])
    #assert nodes point in the correct x direction
    assert (np.sign(wmax - wmin) == np.sign(xmax - xmin))
    #asset span of x points is comperable to span of input x points
    assert (abs(np.var(map[-1]["weight"][:, 0]) - xvar) / xvar < 1.0)
    #assert node points have no (little) y component
    assert (np.var(map[-1]["weight"][:, 1]) < 0.1)
コード例 #9
0
ファイル: test_som.py プロジェクト: dmontemayor/CRPM
def r_test_solve_nestedcs():
    """test nested cs can be solved
    """
    import matplotlib
    matplotlib.use('TkAgg')
    import matplotlib.pyplot as plt

    import numpy as np
    from crpm.setup_nestedcs import setup_nestedcs
    from crpm.fwdprop import fwdprop
    from crpm.som import init_som
    from crpm.som import som
    from crpm.analyzebinaryclassifier import analyzebinaryclassifier

    #init numpy seed
    np.random.seed(40017)

    #setup model
    model, data = setup_nestedcs()

    #update state of model
    _, state = fwdprop(data[0:2, ], model)

    #create and init map with model and its current state
    map, _ = init_som(model, state, n=100, nx=100, ny=1, hcp=True)

    #plot data and map in real space
    plt.scatter(data[0, ], data[1, ], c=data[-1, ])
    plt.plot(map[-1]["weight"][:, 0], map[-1]["weight"][:, 1])
    plt.scatter(map[-1]["weight"][:, 0], map[-1]["weight"][:, 1])
    plt.show()

    #conduct mapping
    pred, map = som(map,
                    state,
                    lstart=.2,
                    lend=.001,
                    nstart=50.0,
                    nend=0.001,
                    maxepoch=5000)

    #plot data and map in real space
    plt.scatter(data[0, ], data[1, ], c=data[-1, ])
    plt.plot(map[-1]["weight"][:, 0], map[-1]["weight"][:, 1])
    plt.scatter(map[-1]["weight"][:, 0], map[-1]["weight"][:, 1])
    plt.show()

    #plot map and centroids in mapping space
    plt.scatter(map[-1]["coord"][:, 0],
                map[-1]["coord"][:, 1],
                c=map[-1]["bias"][:, 0],
                cmap='gray')
    #plt.scatter(map[-1]["centroid"][:,0],map[-1]["centroid"][:,1], s=100)
    plt.show()

    #plot predictions in real space
    #plt.scatter(data[0,], data[1,], c=data[-1,]-pred[0,])
    #plt.show()

    #analyze binary classifier
    _, report = analyzebinaryclassifier(pred, data[-1, ])
    print(report)

    assert report["MatthewsCorrCoef"] >= .5