def test_convsparsenet(lam_sparse=.1,N=16,perc_var=100.): from hdl.models import SparseSlowModel from hdl.learners import SGD whitenpatches = 1000 #model = ConvWhitenInputModel(imshp=(10,1,32,32),convwhitenfiltershp=(7,7),perc_var=100.) model = ConvSparseSlowModel(imshp=(10,1,28,28),convwhitenfiltershp=(7,7),N=N,kshp=(7,7),perc_var=perc_var,lam_sparse=lam_sparse) l = SGD(model=model,datasource='vid075-chunks',display_every=1000,save_every=10000,batchsize=model.imshp[0]) print 'whitenpatches', whitenpatches print 'model.imshp', model.imshp print 'model.convwhitenfiltershp', model.convwhitenfiltershp databatch = l.get_databatch(whitenpatches) l.model.learn_whitening(databatch) l.model.setup() l.learn(iterations=20000) l.change_target(.5) l.learn(iterations=5000) l.change_target(.5) l.learn(iterations=5000) #l.learn(iterations=160000) #l.change_target(.5) #l.learn(iterations=20000) #l.change_target(.5) #l.learn(iterations=20000) from hdl.display import display_final display_final(l.model)
def go(l): databatch = l.get_databatch(whitenpatches) l.model.learn_whitening(databatch) l.model.setup() l.learn(iterations=350000) l.change_target(.5) l.learn(iterations=50000) l.change_target(.5) l.learn(iterations=50000) l.change_target(.5) l.learn(iterations=50000) display_final(l.model)
def test_convsparsenet_subspace(lam_sparse=1.,lam_slow=1.,N=8,perc_var=100.,stride=(1,1)): from hdl.models import ConvSparseSlowModel from hdl.learners import SGD whitenpatches = 1000 psz = 48 ksz = 16 convwhitenfiltershp=(15,15) #model = ConvWhitenInputModel(imshp=(10,1,32,32),convwhitenfiltershp=(7,7),perc_var=100.) model = ConvSparseSlowModel(imshp=(4,1,psz,psz),convwhitenfiltershp=convwhitenfiltershp,N=N,kshp=(ksz,ksz),stride=stride, sparse_cost='subspacel1', perc_var=perc_var, lam_sparse=lam_sparse, lam_slow=lam_slow, mask=True, force_subspace_orthogonal=True) l = SGD(model=model,datasource='vid075-chunks',display_every=50,save_every=10000, eta_target_maxupdate=0.5, batchsize=model.imshp[0]) print 'whitenpatches', whitenpatches print 'model.imshp', model.imshp print 'model.convwhitenfiltershp', model.convwhitenfiltershp databatch = l.get_databatch(whitenpatches) l.model.learn_whitening(databatch) l.model.setup() l.learn(iterations=1000) l.model.center_basis_functions=False l.learn(iterations=9000) l.change_target(.5) l.learn(iterations=5000) l.change_target(.5) l.learn(iterations=5000) #l.learn(iterations=160000) #l.change_target(.5) #l.learn(iterations=20000) #l.change_target(.5) #l.learn(iterations=20000) from hdl.display import display_final display_final(l.model)
def test_convsparsenet(lam_sparse=.1,N=16,perc_var=100.,stride=1,kshp=(7,7),batchsize=4): from hdl.learners import SGD whitenpatches = 1000 if isinstance(kshp,int): kshp = (kshp,kshp) if isinstance(stride,int): stride = (stride,stride) imszr = 5*stride[0] + kshp[0] - 1 imszc = 5*stride[1] + kshp[1] - 1 #model = ConvWhitenInputModel(imshp=(10,1,32,32),convwhitenfiltershp=(7,7),perc_var=100.) model = ConvSparseSlowModel(imshp=(batchsize,1,imszr,imszc),convwhitenfiltershp=(7,7),N=N,kshp=kshp,stride=stride, perc_var=perc_var,lam_sparse=lam_sparse) l = SGD(model=model,datasource='vid075-chunks',display_every=50,save_every=10000,batchsize=model.imshp[0]) print 'whitenpatches', whitenpatches print 'model.imshp', model.imshp print 'model.convwhitenfiltershp', model.convwhitenfiltershp databatch = l.get_databatch(whitenpatches) l.model.learn_whitening(databatch) l.model.setup() l.learn(iterations=5000) l.model.center_basis_functions = False l.learn(iterations=15000) l.change_target(.5) l.learn(iterations=5000) l.change_target(.5) l.learn(iterations=5000) #l.learn(iterations=160000) #l.change_target(.5) #l.learn(iterations=20000) #l.change_target(.5) #l.learn(iterations=20000) from hdl.display import display_final display_final(l.model)
def test_sparsenet_subspace(lam_sparse=1.,lam_slow=1.,N=8,perc_var=100.): from hdl.models import SparseSlowModel from hdl.learners import SGD whitenpatches = 10000 psz = 16 #model = ConvWhitenInputModel(imshp=(10,1,32,32),convwhitenfiltershp=(7,7),perc_var=100.) model = SparseSlowModel(patch_sz=psz,N=N, sparse_cost='subspacel1', perc_var=perc_var, lam_sparse=lam_sparse, lam_slow=lam_slow,T=48) l = SGD(model=model,datasource='vid075-chunks',display_every=100,save_every=10000,batchsize=model.T) print 'whitenpatches', whitenpatches databatch = l.get_databatch(whitenpatches) l.model.learn_whitening(databatch) l.model.setup() l.learn(iterations=5000) l.change_target(.5) l.learn(iterations=5000) l.change_target(.5) l.learn(iterations=5000) #l.learn(iterations=160000) #l.change_target(.5) #l.learn(iterations=20000) #l.change_target(.5) #l.learn(iterations=20000) from hdl.display import display_final display_final(l.model)
l.model.learn_whitening(databatch) newA = np.dot(l.model.whitenmatrix,newA) newA = l.model.normalize_A(newA) l.model.A.set_value(newA.astype(hdl.models.theano.config.floatX)) l.model.reset_functions() return l initial_target = l.eta_target_maxupdate databatch = l.get_databatch(whitenpatches) l.model.learn_whitening(databatch) l.model.setup() l.eta_target_maxupdate iterations = 2000 l = learn_loop(l,iterations=iterations) display_final(l.model,save_string='doubling_0') doublings = 4 for doubling in range(doublings): iterations *= 2 l.eta_target_maxupdate = initial_target print 'Doubling model...' l = double_patch_sz(l) display_final(l.model,save_string='doubling_%d_before_learning'%(doubling+1)) l = learn_loop(l,iterations=iterations) display_final(l.model,save_string='doubling_%d_after_learning'%(doubling+1))
whitenpatches = 400 convwhitenfiltershp=(11,11) perc_var = 99. N = 641 kshp = (16,16) stride = (8,8) imsz = kshp[0]*6 imshp=(2,1,imsz,imsz) print 'Init...' l = SGD(model=ConvSparseSlowModel(imshp=imshp,convwhitenfiltershp=convwhitenfiltershp,perc_var=perc_var,N=N,kshp=kshp,stride=stride, sparse_cost='subspacel1mean',slow_cost=None,lam_sparse=1.,center_basis_functions=False), datasource='berkeleysegmentation',batchsize=imshp[0],save_every=20000,display_every=1000, ipython_profile=profile) print 'Estimate whitening...' databatch = l.get_databatch(whitenpatches) l.model.learn_whitening(databatch) l.model.setup() l.learn(iterations=40000) l.change_target(.5) l.learn(iterations=5000) l.change_target(.5) l.learn(iterations=5000) from hdl.display import display_final display_final(l.model)
#model_name = 'SparseSlowModel_patchsz064_N2048_NN2048_l2_l1_None_2012-02-05_15-29-08/SparseSlowModel_patchsz064_N2048_NN2048_l2_l1_None.model' # faces #model_name = 'SparseSlowModel_patchsz032_N512_NN512_l2_l1_None_2012-02-09_11-40-37/SparseSlowModel_patchsz032_N512_NN512_l2_l1_None.model' #model_name = 'SparseSlowModel_patchsz064_N1024_NN1024_l2_l1_None_2012-02-09_11-47-43/SparseSlowModel_patchsz064_N1024_NN1024_l2_l1_None.model' #model_name = 'SparseSlowModel_patchsz032_N512_NN512_l2_l1_None_2012-02-09_16-38-31/SparseSlowModel_patchsz032_N512_NN512_l2_l1_None.model' #model_name = 'SparseSlowModel_patchsz032_N512_NN512_l2_l1_None_2012-02-09_19-05-45/SparseSlowModel_patchsz032_N512_NN512_l2_l1_None.model' model_name = 'SparseSlowModel_patchsz048_N768_NN768_l2_l1_None_2012-02-09_19-07-07/SparseSlowModel_patchsz048_N768_NN768_l2_l1_None.model' #model_name = 'SparseSlowModel_patchsz064_N1024_NN1024_l2_l1_None_2012-02-09_19-08-50/SparseSlowModel_patchsz064_N1024_NN1024_l2_l1_None.model' model_name = 'SparseSlowModel_patchsz048_N1024_NN1024_l2_l1_None_2012-02-10_16-23-20/SparseSlowModel_patchsz048_N1024_NN1024_l2_l1_None.model' model_name = 'SparseSlowModel_patchsz048_N1024_NN1024_l2_subspacel1_None_2012-02-10_17-22-55/SparseSlowModel_patchsz048_N1024_NN1024_l2_subspacel1_None.model' model_name = 'SparseSlowModel_patchsz048_N4096_NN4096_l2_l1_None_2012-03-01_19-15-33/SparseSlowModel_patchsz048_N4096_NN4096_l2_l1_None.model' model_name = 'SparseSlowModel_patchsz048_N512_NN512_l2_subspacel1_None_2012-03-05_11-42-48/SparseSlowModel_patchsz048_N512_NN512_l2_subspacel1_None.model' #fname = os.path.join(state_dir,model_name) #m = SparseSlowModel() #m.load(fname,reset_theano=False) model_name = 'BinocColorModel_patchsz016_N512_NN512_l2_subspacel1_dist_2012-03-13_15-24-27/BinocColorModel_patchsz016_N512_NN512_l2_subspacel1_dist.model' model_name = 'BinocColorModel_patchsz016_N512_NN512_l2_subspacel1_dist_2012-03-15_15-56-39/BinocColorModel_patchsz016_N512_NN512_l2_subspacel1_dist.model' model_name = 'BinocColorModel_patchsz020_N3200_NN3200_l2_subspacel1_dist_2012-03-18_17-35-13/BinocColorModel_patchsz020_N3200_NN3200_l2_subspacel1_dist.model' fname = os.path.join(state_dir,model_name) m = BinocColorModel() m.load(fname,reset_theano=False) display_final(m) print m