Sparse Modeling in Python

Presentation slides at PyCon PH 2018 Lightning Talks

Transcript

1. Sparse Modeling in Python Feb 25th, 2018 PyCon PH 2018 @tksmd
2. ,ZPUP IUUQTXXXGMJDLSDPNQIPUPTQFESPT[
3. 8IBU
   JT
   z4QBSTF
   .PEFMJOHz • .FUIPE UP BOBMZ[F EBUB
   XJUI
   l4QBSTJUZz • 7BSJPVT EJTDVTTJPOT TUBSUFE
   BSPVOE
   
   JO
   BDBEFNJB • *NBHF
   QSPDFTTJOH
   JT
   POF
   PG
   UIF
   IPU
   BQQMJDBUJPOT
4. "EWBOUBHFT • 4FMFDU
   JNQPSUBOU
   GFBUVSFT
   PG
   JOQVU • 8PSL
   XJUI
   FWFO
   TNBMM
   BNPVOU
   PG
   EBUB
5. 5 %BNBHF
   EFUFDUJPO
   PG
   XBMM
   PG
   CVJMEJOH &YBNQMF
   PG
   JNBHF
   BOBMZTJT
6. #SJFG
   FYBNQMF
   PG
   GFBUVSF
   TFMFDUJPO
7. 1PMZOPNJBM
   3FHSFTTJPO ! = −$%& + $ + Observational Noise
8. 1PMZOPNJBM
   3FHSFTTJPO -JOFBS
   3FHSFTTJPO
   
   0WFSGJUUJOH
   
9. 1PMZOPNJBM
   3FHSFTTJPO $PFGGJDJFOU
   PG
   TJNQMF
   MJOFBS
   SFHSFTTJPO
10. 1PMZOPNJBM
    3FHSFTTJPO -FBTU
    BCTPMVUF
    TISJOLBHF
    BOE
    TFMFDUJPO
    PQFSBUPS
    -BTTP
11. 1PMZOPNJBM
    3FHSFTTJPO $PFGGJDJFOU
    PG
    -"440
12. 1PMZOPNJBM
    3FHSFTTJPO 0SUIPHPOBM
    .BUDIJOH
    1VSTVJU
    0.1
13. 1PMZOPNJBM
    3FHSFTTJPO $PFGGJDJFOU
    PG
    0.1
14. 1PMZOPNJBM
    3FHSFTTJPO from sklearn.linear_model import LinearRegression, Lasso, OrthogonalMatchingPursuit from sklearn.preprocessing import PolynomialFeatures from sklearn.pipeline import make_pipeline poly_preprocess = PolynomialFeatures(poly_dim, include_bias=False) # models linear = LinearRegression() lasso = Lasso(alpha=0.002, max_iter=500000, tol=0.000001) omp = OrthogonalMatchingPursuit(n_nonzero_coefs=5) def fit_and_predict(predictor): model = make_pipeline(poly_preprocess, predictor) model.fit(x.reshape(-1, 1), y) y_predicted = model.predict(x.reshape(-1, 1)) t_predicted = model.predict(t.reshape(-1, 1)) return y_predicted, t_predicted
15. 1PMZOPNJBM
    3FHSFTTJPO • 4UBSU
    XJUI
    MFBTUTRVBSF
    NFUIPE min 1 2 & − () * Z  PVUQVU X 
    XFJHIU Y 
    JOQVU $PNQVUF X UP
    TBUJTGZ
    BCPWF
16. 1PMZOPNJBM
    3FHSFTTJPO • "EE
    SFHVMBUJPO
    UFSN
    UP
    BWPJE
    PWFSGJUUJOH min 1 2 & − () * + , ( - 4VSQSFTT PWFSGJUUJOH CZ BEEJOH DPOTUSBJOU UP XFJHIU
    X -
    /PSN -BTTPɾ-
    /PSN
    3JEHF
17. 1PMZOPNJBM
    3FHSFTTJPO • "QQSPBDI
    UP
    -
    /PSN
    0QUJNJ[BUJPO min 1 2 & − () * + , ( - 5IJT JT
    FTTFOUJBMMZ
    DPNCJOBUJPOBM
    PQUJNJ[BUJPO
    QSPCMFN /1IBSE (SFFEZ
    BMHPSJUIN
    UP
    TPMWF
    JU
    MPDBMMZ
    MJLF
    .BUDIJOH
    1VSTVJU
    *)5
18. 4VNNBSZ • *OUSPEVDUJPO
    PG
    4QBSTF
    .PEFMJOH
    JO
    1ZUIPO • 4PNF
    JNQMFNFOUBUJPO
    JT
    BMSFBEZ
    BWBJMBCMF
    JO
    /VN1Z PS
    TDJLJUMFBSO • +VQZUFS OPUFCPPL
    JT
    BWBJMBCMF
    CFMPX • IUUQTHJUJPW/&O
19. -PDBM
    6TFS
    (SPVQ
    JO
    ,ZPUP IUUQTIBOOBSJQZUIPODPOOQBTTDPN SE 'SJEBZ
    FWFSZ
    NPOUI