Python io.mmread函数代码示例

def eigenbase (h1, d1, E, v0, pow0, pow1, rest):
  # compute all eigenvalues and eigenvectors
  pt0 = 'out/impacting-bar/MK_%g_%g_%g_%g_%d_%d'%(h1, d1, E, v0, pow0, pow1)
  sl0 = SOLFEC ('DYNAMIC', 1E-3, pt0)
  bl0 = BULK_MATERIAL (sl0, model = 'KIRCHHOFF', young = E, poisson = PoissonRatio, density = MassDensity)
  bod = BODY (sl0, 'FINITE_ELEMENT', COPY (mesh), bl0)
  eval = [] # selected eigenvalue list
  evec = [] # selected eigenvector list (BODY command takes a tuple (eval, evec) argument for the RO formulation)
  vsel = (0,1,2,3,4,5,13,18,25,33,38)
  if 0:
    BODY_MM_EXPORT (bod, pt0+'/M.mtx', pt0+'/K.mtx')
    M = mmread (pt0+'/M.mtx').todense()
    K = mmread (pt0+'/K.mtx').todense()
    for j in range (0, K.shape[1]):
      for i in range (j+1, K.shape[0]):
	K [j, i] = K [i, j] # above diagonal = below diagonal
    x, y = eigh (K, M) # this produces y.T M y = 1 and y.T K y = x */
    for j in vsel:
      eval.append (x[j].real)
      for z in y[:,j]:
	evec.append (z.real)
  else:
    data0 = MODAL_ANALYSIS (bod, 45, pt0 + '/modal.data', verbose = 'ON', abstol = 1E-14)
    ndofs = mesh.nnod * 3
    for j in vsel:
      eval.append (data0[0][j])
      for k in range (j*ndofs,(j+1)*ndofs):
	evec.append (data0[1][k])
  return (eval, evec)

def ro0_modal_base (use_scipy=False, verbose='OFF'):
  sol = ro0_model (1E-3, 0.0)
  bod = sol.bodies[0]
  eval = [] # selected eigenvalue list
  evec = [] # selected eigenvector list
  vsel = (0,1,2,3,4,5,13,18,25,33,38)
  if use_scipy:
    BODY_MM_EXPORT (bod, 'out/reduced-order0/M.mtx',
                         'out/reduced-order0/K.mtx')
    M = mmread ('out/reduced-order0/M.mtx').todense()
    K = mmread ('out/reduced-order0/K.mtx').todense()
    for j in range (0, K.shape[1]):
      for i in range (j+1, K.shape[0]):
	K [j, i] = K [i, j] # above diagonal = below diagonal
    x, y = eigh (K, M) # this produces y.T M y = 1 and y.T K y = x
    for j in vsel:
      eval.append (x[j].real)
      for z in y[:,j]:
	evec.append (z.real)
  else:
    data0 = MODAL_ANALYSIS (bod, 45, 'out/reduced-order0/modal',
                            1E-13, 1000, verbose)
    dofs = len(bod.velo)
    for j in vsel:
      eval.append (data0[0][j])
      for k in range (j*dofs,(j+1)*dofs):
	evec.append (data0[1][k])

  return (eval, evec)

def applySVMWithPCA():
    '''
    Same as the previous function, just change the file names..
    '''
    data = io.mmread(ROOTDIR+"TRAINDATA.mtx")
    label = np.load(ROOTDIR+"label_train.npy")
    testdata = io.mmread(ROOTDIR+"TESTDATA.mtx")
    testLabel = np.load(ROOTDIR + "label_test.npy")
    
    linear_svm = LinearSVC(C=1.0, class_weight=None, loss='hinge', dual=True, fit_intercept=True,
    intercept_scaling=1, multi_class='ovr', penalty='l2',
    random_state=None, tol=0.0001, verbose=1, max_iter=2000)
     
    data = scale(data, with_mean=False)
     
    linear_svm.fit(data, label)
    joblib.dump(linear_svm, ROOTDIR+'originalTrain_hinge_2000.pkl') 
#     linear_svm = joblib.load(ROOTDIR+'originalTrain_hinge_2000.pkl')
    
    print 'Trainning Done!'
    scr = linear_svm.score(data, label)
    print 'accuracy on the training set is:' + str(scr)

    predLabel = linear_svm.predict(data)
    calcualteRMSE(label, predLabel)
    
    scr = linear_svm.score(testdata, testLabel)
    print 'accuracy on the testing set is:' + str(scr)

    predLabel = linear_svm.predict(testdata)
    calcualteRMSE(testLabel, predLabel)      

def read_input_tensor(headers_filename, data_file_names, tensor_slices, adjustDim=False, offerString="Attr: OFFER",
                      wantString="Attr: WANT"):

    #load the header file
    _log.info("Read header input file: " + headers_filename)
    input = codecs.open(headers_filename,'r',encoding='utf8')
    headers = input.read().splitlines()
    input.close()

    # get the largest dimension of all slices
    if adjustDim:
        maxDim = 0
        for data_file in data_file_names:
            matrix = mmread(data_file)
            if maxDim < matrix.shape[0]:
                maxDim = matrix.shape[0]
            if maxDim < matrix.shape[1]:
                maxDim = matrix.shape[1]

    # load the data files
    slice = 0
    tensor = SparseTensor(headers, offerString, wantString)
    for data_file in data_file_names:
        if adjustDim:
            adjusted = adjust_mm_dimension(data_file, maxDim)
            if adjusted:
                _log.warn("Adujst dimension to (%d,%d) of matrix file: %s" % (maxDim, maxDim, data_file))
        _log.info("Read as slice %d the data input file: %s" % (slice, data_file))
        matrix = mmread(data_file)
        tensor.addSliceMatrix(matrix, tensor_slices[slice])
        slice = slice + 1
    return tensor

def get_debug(data):
    full_train = sio.mmread('data/%s_train.mtx' % data).tocsr()
    (nu, nm) = full_train.shape

    print 'sampling'
    debug_mids = sample(range(nm), nm / 5)
    debug_uids = sample(range(nu), nu / 5)

    debug = full_train[debug_uids][:, debug_mids].tocoo()
    nr = debug.nnz
    train_ids, _, test_ids = sample_split(nr)

    # build matrix from given indices
    print 'writing debug_train'
    debug_train = coo_matrix(
        (debug.data[train_ids], (debug.row[train_ids], debug.col[train_ids])), debug.shape)
    sio.mmwrite('data/%s_debug_train.mtx' % data, debug_train)
    print 'writing debug_test'
    debug_test = coo_matrix(
        (debug.data[test_ids], (debug.row[test_ids], debug.col[test_ids])), debug.shape)
    sio.mmwrite('data/%s_debug_test.mtx' % data, debug_test)

    # build movie mtx from debug_mids
    print 'movie debug'
    movies = sio.mmread('data/movies.mtx').tocsr()
    movies_debug = movies[debug_mids]
    sio.mmwrite('data/movies_%s_debug.mtx' % data, movies_debug)

    return debug, debug_train, debug_test, movies_debug

def load(ppt, samples, l_tau, l_lc, l_regtype, b_tau, b_lc, b_regtype):

    ln = np.loadtxt('lin-models/bestlinwtln'+l_regtype+samples+'tau'+l_tau+'lc'+l_lc+ppt+'.txt')
    lv = np.loadtxt('lin-models/bestlinwtlv'+l_regtype+samples+'tau'+l_tau+'lc'+l_lc+ppt+'.txt')
    bv = np.loadtxt('bil-models/bestbilwtbn'+b_regtype+samples+'tau'+b_tau+'eta'+b_lc+ppt+'.txt')
    bn = np.loadtxt('bil-models/bestbilwtbv'+b_regtype+samples+'tau'+b_tau+'eta'+b_lc+ppt+'.txt')

    traindata = [(d.strip().split()[1:5], d.strip().split()[5]) for d in open('clean/cleantrain.txt')]
    devdata = [(d.strip().split()[1:5], d.strip().split()[5]) for d in open('clean/cleandev.txt')]
    testdata = [(d.strip().split()[1:5], d.strip().split()[5]) for d in open('clean/cleantest.txt')]

    traindata = traindata[:int(samples)]

    phih = sio.mmread('clean/trh1k.mtx')
    phim = sio.mmread('clean/trm1k.mtx')
    phidh = sio.mmread('clean/devh1k.mtx')
    phidm = sio.mmread('clean/devm1k.mtx')
    maph = np.loadtxt('clean/forhead.txt', dtype=str)
    mapm = np.loadtxt('clean/formod.txt', dtype=str)
    mapdh = np.loadtxt('clean/devheads.txt', dtype=str)
    mapdm = np.loadtxt('clean/devmods.txt', dtype=str)


    trainingdat = bilme.BilinearMaxentFeatEncoding.train(traindata, phih, phim, maph, mapm, pptype=ppt)
    traintoks = trainingdat.train_toks()
    traintokens = [(co.word_features(t),l) for t,l in trainingdat.train_toks()]
    devencode = bilme.BilinearMaxentFeatEncoding.train(devdata, phidh, phidm, mapdh, mapdm, pptype=ppt)
    devtoks = devencode.train_toks()
    devtokens = [(co.word_features(t),l) for t,l in devencode.train_toks()]

    data = [devtoks, devtokens]

    trlinencoding = maxent.BinaryMaxentFeatureEncoding.train(traintokens)

    return trlinencoding, devencode, [ln, lv], [bn, bv], data

def generate_valid_repos_and_times(dataset_dir):
    """Function called to generate VALID_REPOS_AND_TIMES in `dataset_dir`
    """
    valid_repos_and_times = []

    repos_users_times_fn = join(dataset_dir, TIMED_INTERESTS_FN)
    u_r_t = mmread(repos_users_times_fn).transpose().tocsr()

    validation_repos_fn = join(dataset_dir, VALIDATING_FN)
    validation_matrix = mmread(validation_repos_fn).tocsr()

    v_u_r_t = u_r_t.multiply(validation_matrix).tolil()

    for uidx in xrange(v_u_r_t.shape[0]):
        v_r_t_coo = v_u_r_t.getrowview(uidx).tocoo()
        sorted_index = np.argsort(v_r_t_coo.data)

        times = v_r_t_coo.data[sorted_index]
        repos = v_r_t_coo.col[sorted_index]
        valid_repos_and_times.append(np.vstack((times,repos)))

    pt_fn = join(dataset_dir, VALID_REPOS_AND_TIMES)
    with open(pt_fn, "wb") as pf:
        cPickle.dump(valid_repos_and_times, pf, cPickle.HIGHEST_PROTOCOL)
    return pt_fn

def main():

    import os
    import logging
    import subprocess
    from optparse import OptionParser
    import numpy as np
    from scipy.io import mmread

    from mrec import save_recommender
    from mrec.mf.recommender import MatrixFactorizationRecommender
    from filename_conventions import get_modelfile

    logging.basicConfig(level=logging.INFO,format='[%(asctime)s] %(levelname)s: %(message)s')

    parser = OptionParser()
    parser.add_option('--factor_format',dest='factor_format',help='format of factor files tsv | mm (matrixmarket) | npy (numpy array)')
    parser.add_option('--user_factors',dest='user_factors',help='user factors filepath')
    parser.add_option('--item_factors',dest='item_factors',help='item factors filepath')
    parser.add_option('--train',dest='train',help='filepath to training data, just used to apply naming convention to output model saved here')
    parser.add_option('--outdir',dest='outdir',help='directory for output')
    parser.add_option('--description',dest='description',help='optional description of how factors were computed, will be saved with model so it can be output with evaluation results')

    (opts,args) = parser.parse_args()
    if not opts.factor_format or not opts.user_factors or not opts.item_factors \
            or not opts.outdir:
        parser.print_help()
        raise SystemExit

    model = MatrixFactorizationRecommender()

    logging.info('loading factors...')

    if opts.factor_format == 'npy':
        model.U = np.load(opts.user_factors)
        model.V = np.load(opts.item_factors)
    elif opts.factor_format == 'mm':
        model.U = mmread(opts.user_factors)
        model.V = mmread(opts.item_factors)
    elif opts.factor_format == 'tsv':
        model.U = np.loadtxt(opts.user_factors)
        model.V = np.loadtxt(opts.item_factors)
    else:
        raise ValueError('unknown factor format: {0}'.format(factor_format))

    if opts.description:
        model.description = opts.description

    logging.info('saving model...')

    logging.info('creating output directory {0}...'.format(opts.outdir))
    subprocess.check_call(['mkdir','-p',opts.outdir])

    modelfile = get_modelfile(opts.train,opts.outdir)
    save_recommender(model,modelfile)

    logging.info('done')

def fit_lightfm_model():
	""" Fit the lightFM model 
	
	returns d_user_pred, list_user, list_coupon
	list_coupon = list of test coupons 
	list_user = list of user ID 
	d_user_pred : key = user, value = predicted ranking of coupons in list_coupon
	"""

	#Load data
	Mui_train = spi.mmread("../Data/Data_translated/biclass_user_item_train_mtrx.mtx")
	uf        = spi.mmread("../Data/Data_translated/user_feat_mtrx.mtx")
	itrf      = spi.mmread("../Data/Data_translated/train_item_feat_mtrx.mtx")
	itef      = spi.mmread("../Data/Data_translated/test_item_feat_mtrx.mtx")
	
	#Print shapes as a check
	print "user_features shape: %s,\nitem train features shape: %s,\nitem test features shape: %s"   % (uf.shape, itrf.shape, itef.shape)
	
	#Load test coupon  and user lists
	cplte       = pd.read_csv("../Data/Data_translated/coupon_list_test_translated.csv")
	ulist       = pd.read_csv("../Data/Data_translated/user_list_translated.csv")
	list_coupon = cplte["COUPON_ID_hash"].values
	list_user   = ulist["USER_ID_hash"].values
	
	#Build model
	no_comp, lr, ep = 10, 0.01, 5
	model = LightFM(no_components=no_comp, learning_rate=lr, loss='warp')
	model.fit_partial(Mui_train, user_features = uf, item_features = itrf, epochs = ep, num_threads = 4, verbose = True)

	test               = sps.csr_matrix((len(list_user), len(list_coupon)), dtype = np.int32)
	no_users, no_items = test.shape
	pid_array          = np.arange(no_items, dtype=np.int32)

	#Create and initialise dict to store predictions
	d_user_pred = {}
	for user in list_user :
		d_user_pred[user] = []
	
	# Loop over users and compute predictions
	for user_id, row in enumerate(test):
		sys.stdout.write("\rProcessing user " + str(user_id)+"/ "+str(len(list_user)))
		sys.stdout.flush()
		uid_array         = np.empty(no_items, dtype=np.int32)
		uid_array.fill(user_id)
		predictions       = model.predict(uid_array, pid_array,user_features = uf, item_features = itef, num_threads=4)
		user              = str(list_user[user_id])
		# apply MinMaxScaler for blending later on
		MMS               = MinMaxScaler()
		pred              = MMS.fit_transform(np.ravel(predictions))
		d_user_pred[user] = pred

	# Pickle the predictions for future_use
	d_pred = {"list_coupon" : list_coupon.tolist(), "d_user_pred" : d_user_pred}
	with open("../Data/Data_translated/d_pred_lightfm.pickle", "w") as f:
		pickle.dump(d_pred, f, protocol = pickle.HIGHEST_PROTOCOL)

	return d_user_pred, list_user, list_coupon

def goMusic(K=80,steps=200,resume=False,normalize=True,R=None,V=None,mean_center=False,beta=0.0,betaO=0.0,normalizer=normalizer,doBias=True,every=1,doFactors=True,biasSteps=10):
    #R = mmread("reviews_Musical_Instruments.mtx").tocsr()
    if R == None:
        R = mmread("training.mtx").tocsr().toarray()
    else:
        R = R.toarray()

    if V == None:
        V = mmread("validation.mtx").todok()
    
    mu = np.finfo(float).eps



    if normalize:
        R = normalizer(R,1,0)
        print "normalizing, min/max", R.min(),R.max()

    
    #R = R[0:424,:]
    if not resume:
        P = normalizer(np.random.rand(R.shape[0],K),.1,0)
        Q = normalizer(np.asfortranarray(np.random.rand(K,R.shape[1])),.1,0)

        #bP,bQ = makeAvgBaseline(R)
        #print bP,bQ
        bP = None # np.zeros(R.shape[0])#None
        bQ = None #np.zeros(R.shape[1])#None#(R > 0).mean(axis=0)
        #bP,bQ = makeAvgBaseline(R)
    else:
        P = np.loadtxt("P.txt")
        Q = np.loadtxt("Q.txt")
        bP = np.loadtxt("bP.txt")
        bQ = np.loadtxt("bQ.txt")

    print R.shape,P.shape,Q.shape
    print "starting doFactO"
    #chunkFactO(R,P,Q,K,steps=steps,chunks=1,discard=0)#chunks=800,discard=0)

    #R,P,Q,bP,bQ = factO(R,P,Q,K,steps=steps,discard=0,bP=bP,bQ=bQ,beta=beta,betaO=betaO)
    rmses,maes,errs = [],[],[]

    def validation(P,Q,bP,bQ):
        rmse,mae,err = validate(T=R,V=V,P=P,Q=Q,bP=bP,bQ=bQ)
        rmses.append(rmse)
        maes.append(mae)
        errs.append(err)

    R,P,Q,bP,bQ,t_rmses = sigFactO(R,P,Q,K,bP=bP,bQ=bQ,steps=steps,discard=0.0,beta=beta,betaO=betaO,mean_center=mean_center,doBias=doBias,validate=validation,every=every,doFactors=doFactors,biasSteps=biasSteps)    

    if normalize:
        R = renormalizer(R,1,0,5,0)

    dumparrays(R,P,Q,bP,bQ)


    return t_rmses,rmses,maes,errs

def main():
    train_tfidf = sio.mmread(tfidf_train_file)
    test_tfidf = sio.mmread(tfidf_test_file)

    svd = TruncatedSVD(400)
    svd_X_train = svd.fit_transform(train_tfidf)
    svd_X_test = svd.transform(test_tfidf)

    sio.mmwrite('train_tfidf_2013_svd_400_mtx', svd_X_train)
    sio.mmwrite('test_tfidf_svd_400_mtx', svd_X_test)

def create_tox21(sparsity_cutoff, validation_fold, dtype=np.float32, download_directory=_DATA_DIRECTORY):
    urlbase = "http://www.bioinf.jku.at/research/deeptox/"
    dst = os.path.join(download_directory, "raw")
    fn_x_tr_d = _download_file(urlbase, "tox21_dense_train.csv.gz", dst)
    fn_x_tr_s = _download_file(urlbase, "tox21_sparse_train.mtx.gz", dst)
    fn_y_tr = _download_file(urlbase, "tox21_labels_train.csv", dst)
    fn_x_te_d = _download_file(urlbase, "tox21_dense_test.csv.gz", dst)
    fn_x_te_s = _download_file(urlbase, "tox21_sparse_test.mtx.gz", dst)
    fn_y_te = _download_file(urlbase, "tox21_labels_test.csv", dst)
    cpd = _download_file(urlbase, "tox21_compoundData.csv", dst)

    y_tr = pd.read_csv(fn_y_tr, index_col=0)
    y_te = pd.read_csv(fn_y_te, index_col=0)
    x_tr_dense = pd.read_csv(fn_x_tr_d, index_col=0).values
    x_te_dense = pd.read_csv(fn_x_te_d, index_col=0).values
    x_tr_sparse = io.mmread(fn_x_tr_s).tocsc()
    x_te_sparse = io.mmread(fn_x_te_s).tocsc()

    # filter out very sparse features
    sparse_col_idx = ((x_tr_sparse > 0).mean(0) >= sparsity_cutoff).A.ravel()
    x_tr_sparse = x_tr_sparse[:, sparse_col_idx].A
    x_te_sparse = x_te_sparse[:, sparse_col_idx].A

    dense_col_idx = np.where(x_tr_dense.var(0) > 1e-6)[0]
    x_tr_dense = x_tr_dense[:, dense_col_idx]
    x_te_dense = x_te_dense[:, dense_col_idx]

    # The validation set consists of those samples with
    # cross validation fold #5
    info = pd.read_csv(cpd, index_col=0)
    f = info.CVfold[info.set != "test"].values
    idx_va = f == float(validation_fold)

    # normalize features
    from sklearn.preprocessing import StandardScaler

    s = StandardScaler()
    s.fit(x_tr_dense[~idx_va])
    x_tr_dense = s.transform(x_tr_dense)
    x_te_dense = s.transform(x_te_dense)

    x_tr_sparse = np.tanh(x_tr_sparse)
    x_te_sparse = np.tanh(x_te_sparse)

    x_tr = np.hstack([x_tr_dense, x_tr_sparse])
    x_te = np.hstack([x_te_dense, x_te_sparse])

    return (
        x_tr[~idx_va].astype(dtype, order="C"),
        y_tr[~idx_va].values.astype(dtype, order="C"),
        x_tr[idx_va].astype(dtype, order="C"),
        y_tr[idx_va].values.astype(dtype, order="C"),
        x_te.astype(dtype, order="C"),
        y_te.values.astype(dtype, order="C"),
    )

def main():
    FORMAT = '%(asctime)s %(levelname)s %(message)s'
    logging.basicConfig(format=FORMAT)
    logging.getLogger().setLevel(logging.INFO)
    args = parse_args()
    lang_map = {i: fn for i, fn in enumerate(sorted(listdir(args.lang_map)))}
    if args.train.endswith('.mtx'):
        mtx = mmread(args.train).todense()
        t_mtx = mmread(args.test).todense()
    else:
        with open(args.train) as stream:
            mtx = np.loadtxt(stream, np.float64)
        with open(args.test) as stream:
            t_mtx = np.loadtxt(stream, np.float64)
    labels = np.ravel(mtx[:, 0])
    test_labels = t_mtx[:, 0]
    test_mtx = t_mtx[:, 1:]
    if args.scale:
        train = scale(mtx[:, 1:], with_mean=False)
    else:
        train = mtx[:, 1:]
    kwargs = {}
    for a in args.params:
        k, v = a.split('=')
        try:
            v = int(v)
        except:
            pass
        kwargs[k] = v
    r = Representation(args.encoder, args.classifier, **kwargs)
    r.encode(train)
    logging.info('Matrix encoded')
    r.train_classifier(labels)
    logging.info('Model trained')
    acc = 0
    N = 0
    for vec_ in test_mtx:
        vec = np.ravel(vec_)
        cl = r.classify_vector(vec, with_probs=args.with_probs)
        try:
            lab = test_labels[N, 0]
        except IndexError:
            lab = test_labels[N]
        N += 1
        if args.with_probs:
            guess = max(enumerate(cl[0, :]), key=lambda x: x[1])[0]
            print('{0}\t{1}\t{2}'.format('\t'.join(map(str, cl[0, :])), lang_map[guess], lang_map[int(lab)]))
        else:
            try:
                guess = int(cl[0, 0])
            except IndexError:
                guess = int(cl + 0.5)
            print('{0}\t{1}'.format(lang_map[guess], lang_map[int(lab)]))
        if int(guess) == int(lab):
            acc += 1

def create_bars (h1, E, frict, damp, formulation):

  # compute all eigenvalues and eigenvectors
  if formulation == 'RO':
    pt0 = 'out/16-bars/MK_%g_%g_%g_%g'%(h1, E, frict, damp)
    sl0 = SOLFEC ('DYNAMIC', 1E-3, pt0)
    bl0 = BULK_MATERIAL (sl0, model = 'KIRCHHOFF', young = E, poisson = PoissonRatio, density = MassDensity)
    bod = BODY (sl0, 'FINITE_ELEMENT', COPY (mesh), bl0)
    eval = [] # selected eigenvalue list
    evec = [] # selected eigenvector list (BODY command takes a tuple (eval, evec) argument for the RO formulation)
    vsel = range (0, 32)

    if 0:
      BODY_MM_EXPORT (bod, pt0+'/M.mtx', pt0+'/K.mtx')
      M = mmread (pt0+'/M.mtx').todense()
      K = mmread (pt0+'/K.mtx').todense()
      for j in range (0, K.shape[1]):
	for i in range (j+1, K.shape[0]):
	  K [j, i] = K [i, j] # above diagonal = below diagonal
      x, y = eigh (K, M) # this produces y.T M y = 1 and y.T K y = x */
      for j in vsel:
	eval.append (x[j].real)
	for z in y[:,j]:
	  evec.append (z.real)
    else:
      data0 = MODAL_ANALYSIS (bod, 45, pt0 + '/modal.data', verbose = 'ON', abstol = 1E-14)
      ndofs = mesh.nnod * 3
      for j in vsel:
	eval.append (data0[0][j])
	for k in range (j*ndofs,(j+1)*ndofs):
	  evec.append (data0[1][k])
    data = (eval, evec)

  # 16 bars domain
  sl2 = SOLFEC ('DYNAMIC', h1, 'out/16-bars/%s_%g_%g_%g_%g'%(formulation, h1, E, frict, damp))
  SURFACE_MATERIAL (sl2, model = 'SIGNORINI_COULOMB', friction = frict, restitution = 0.0)
  bl2 = BULK_MATERIAL (sl2, model = 'KIRCHHOFF', young = E, poisson = PoissonRatio, density = MassDensity)
  GRAVITY (sl2, (0, 0, -9.8))
  for i in range (0, nw):
    for j in range (0, nw):
      shp = COPY (mesh)
      TRANSLATE (shp, ((1-nw)*0.05+0.1*i, (1-nw)*0.05+0.1*j, 0))
      if formulation == 'RO':
	bd2 = BODY (sl2, 'FINITE_ELEMENT', shp, bl2, form = formulation, modal = data)
	bd2.scheme = 'DEF_LIM'
	bd2.damping = damp
      elif formulation == 'BC':
	bd2 = BODY (sl2, 'FINITE_ELEMENT', shp, bl2, form = formulation)
	bd2.scheme = 'DEF_LIM'
	bd2.damping = damp
      else: bd2 = BODY (sl2, 'RIGID', shp, bl2)
  BODY (sl2, 'OBSTACLE', COPY (obsm), bl2)

  return sl2

def validate(trunc = False,T = None,V = None,doRound=False,activation=sigmoid,P=None,Q=None,bP=None,bQ=None):
    if T == None:
        Rtraining = mmread('training.mtx').tocsr()
    else:
        Rtraining = T

    if V == None:
        R = mmread('validation.mtx').todok()
    else:
        R = V.todok()
    mean = (Rtraining.sum()) / (Rtraining > 0).sum()
    if not (P != None or Q != None or bP != None or bQ != None):
        P,Q,bP,bQ = np.loadtxt("P.txt"),np.loadtxt("Q.txt"),np.loadtxt("bP.txt"),np.loadtxt("bQ.txt")

    print R.shape,P.shape,Q.shape
    i = 0
    sum = 0
    sumAbs = 0
    lte1 = 0
    sumlte1 = 0
    errors = []
    for k,v in R.items():
        g = bP[k[0]] + bQ[k[1]] + np.dot(P[k[0],:],Q[:,k[1]]) 
        #if trunc:
        #    g = min(1,max(5,g))
        #for i in xrange(P.shape[1]):
        #    g += (P[k[0],i]) * (Q[i,k[1]])
        #    
        #    if trunc:
        #        g = max(1,min(g,5))
        g = activation(mean + g)
        g = renormalizefloat(g,1,0,5,0)

        
        if doRound:
            g = round(g)
        e = (v - g)**2
        sumAbs += math.sqrt((v - g)**2)
        errors.append(e)
        if e < 1.00001:
            lte1 += 1
            sumlte1 += e
        sum += e
        #if e > 5:
        #print i,v,g,e
        i+=1
    rmse = math.sqrt(sum/R.nnz)
    mae = sumAbs / R.nnz
    print "rmse",rmse
    print "mae",sumAbs / R.nnz
    print "lte1",lte1,len(R.items()), lte1/float(len(R.items()))
    print "lte1 rmse",math.sqrt((sumlte1 +1) / (lte1+1))
    print "validation mean",mean
    return rmse,mae,np.array(errors)