RWC 10 RWC 1992 4 2001 13 RWC 21 RWC 1. RWC (Real World Computing ) (1982 1992 ) 3 1992 2001 10 21 RWCP 5 1992 1996 5 1997 2001 RWC (RWI) (PDC) 2 RWC (5G) 1
1) (AI) 5G AI AI Google IT Deep Learning RWC (RWCP) RWC Web RWCP [1] 2. RWC 2.. 1 1971 4 1970 7 ETL-Mark I, II (1952, 1955) (ETL) 1954 1965 (ETL-Mark III VI) ( ) ( [2][3]) (1966 1972) 1) RWC 2
OCR ASPET/70, 71 2) 60 (PIPS) : 1971 1980 TSS 1972 11 1973 3) 10 Lie [4] [5] (HLAC) 1981 [6] 10 Bayes 2 Boole 1 4) 70 (AI) (If-then rule) (KE) 2 AI KE 1969 ARPA-net 70 2) 3) 4) A 3
1: 80 TCP/IP 1980 5) AI KE AI (5G) (1982 1992) ICOT PIM OS (1981 1988) 1985 1992 1982 15 NRC [8] [11] From Boolean to Bayesian 6) [9][10] 80 2 AI KE 5G PIM 70 5) 6) Zadeh 1994 Soft Computing [23] 4
( 2 ) 2: 2.. 2 3 3: AI 5
[37] AI 1956 McCarthy Minsky Artificial Intelligence Newell AI Logic Theorist 60 1965 ELIZA 1966 1969 AI 7) McCarthy, Hayes 1972 NP AI AI 70 AI MYCIN DENDRAL Feigenbaum AI KE 80 80 1982 10 AI If A, then B AI McCulloch-Pitts, 1943 Hebb, 1949 Perceptron Rosenblatt, 1957 7) AI 6
Widrow-Hoff, 1960, 1967 1968 Minsky Papert Perceptrons 80 Hinton, 1984 Hopfield, 1985 Rumelhart, 1986 AI PDP (Parallel Distributed Processing) Pearl, 1988 90 Chow, 1957 1971 10 60 Widrow-Hoff Fisher Koford-Groner, 1966 Patterson-Womack, 1966 Fisher 1972, 1981 [6] A : 80 AI 7
AI 80 Toy world Real world Well-defined Ill-defined 3. RWC 1988 5G 10 5 8
RWC 3.. 1 1989 21 ( ) [12] 1991 [13] NIPT (New Information Processing Technology) RWC (Real World Computing) 1991 8) 3.. 2 [12, 13, 15, 17, 18, 21] 21 8) 5G 3D [16] 9
/ / subsymbolic 3.. 3 : : : 3 4 10
4: 1 [14, 15, 16, 17, 18, 19, 21, 22] 5G 5G 3.. 4 1992-1996 11
1997-2001 GA 9) RWC RWC 5 5: RWC [13] ( ) 9) Genetic Algorithm : 12
4. RWC 1992 7 RWC (RWCP: RWC Partnership) 10 (TRC) TRC RWC RWC TRC TRC RWC 4.. 1 1992 1996, H4 H8 RWCP 12 2 3 23 37 12 4 6 3 12 GMD ISS SNN SICS 4 7 (GMD 2001 10) 1993 5 (ISS KRDL ) (SNN) (SICS) 3 36 50 57 1 RWC NIPT 10) 67 24,000 10 13
1: [35] RWCP H4 H5 H6 H7 H8 TRC OS NTT MRI NEC NEC MRI NEC LD OBIS ISS SNN SICS GMD GMD SICS ISS 14
11) JOP RWC 2 34 RISC-LINZ IRST 1994 [35] ETL RWC RWCP TRC 2 RWC TRC 10 60 RWC 2 [20] 2: ETL-RWC H4 H5 H6 H7 H8 EM-X OS 1994 6 RWCP RWCP Joint Symposium 500 RWC 11) NIPT AI 15
[24] [25] [35] RWC RWCP 12) TRC 1994 3 RWC 1995 4 TRC RWC-1 OB RWC [3] RWC 4.. 2 1996, H8 1996 TRC RWC RWI PDC 12) 16
RWC 13) TRC [28] [31] RWIC 1997 RWC 4.. 3 1997 2001, H9 H13 TRC RWIC 6 RWI RWCP [31][32][33] 6: [35] 13) TRC 17
RWI 7 [29][31] 7: [35] RWIC RWC RWCP TRC 18
PDC 14) 8 [30][31] 8: PDC [35] TRC 3 RWIC 4 RWCP 3: RWI RWI RWC 14) 2000 H12 NEDO 19
4: RWCP RWCP RWI TRC Cross Madiator Cross Mediator NTT MRI NEC NEC ISS KRDL ISS KRDL SICS SNN SNN GMD SICS RWCP PDC TRC PAPIA NEC NEC LAN LD MRI GMD 20
RWI PDC TRC PC TRC RWI 19 PDC 11 [35] 5. [21] [18] [19] RWC 92 [16] RWC 94, 95, 97, 98, 00, 01 RWCP RWC NEWS [1] 1995 4 RWIC [38] [35] 5.. 1 RWC RWC 97 1997 1 600 15) (GA) EM-X RWC-1 PC RWC-1 RWC RWC NEWS [26] RWC [27] [28] RWI PDC TRC [29][30] 15) SF HAL 9000 21
5.. 2 [1] [35] RWI (Jijo2) BAYONET CrossMediator PDC RHiNET SCore Omni OpenMP PAPIA 2001 10 3 5 RWC2001 600 2,000 RWC NEWS Vol. 20 ( ) [34] 10 RWC [35] [36] B RWCP RWIC 2002 3 PWC 5.. 3 TRC RWI RWC RWC 6 22
RWC 21 COE [40] 2001 RWC 13 HLAC CHLAC 16) RWC 2002 2005 2006 TRC RWCP PC 1024 cpu RWC RWI RWC AI Deep Learnig Google IT AI RWI 15 2015 17) AI RWI [39] RWC-RWI 6. RWC ICT iphone 16) RWC 17) RWC 23
RWC AI RWC RWI ICT RWC NEDO 18) RWC AI RWC RWC Web RWCP [1] RWCP RWI 18) RWC 24
[1] RWCP, 10 [13, 31, 35, 36] RWC NEWS. (http://keima.la.coocan.jp/rwcp/memorial/index.html) [2],. (http://museum.ipsj.or.jp/computer/dawn/index.html) [3] :,, IPSJ Magazine, Vol. 44, No. 10 (Oct. 2003). (http://museum.ipsj.or.jp/guide/pdf/magazine/ipsj-mgn441015.pdf) [4] N. Otsu: An Invariant Theory of Linear Functionals as Linear Feature Extractors, Bull. ETL, Vol. 37, No. 10, pp. 893 913 (1973). [5] N. Otsu: Nonlinear Discriminant Analysis as a Natural Extension of the Linear Case, Behaviormetrika, Vol. 2, pp. 45 59 (1975). [6] :,, 818, 210 (1981). [7] N. Otsu: Optimal Linear and Nonlinear Solutions for Least-square Discriminant Feature Extraction, Proc. of 6th Int. Conf. on Pattern Recognition, pp. 557 560 (1982). [8],,,,, :,, 211, 136 (1985). [9] :,, Vol. 71, No. 11, pp. 1231 1240 (1988). [10] N. Otsu: Toward Soft Logic for the Foundation of Felxible Information Processing, Bull. ETL, Vol. 53, No. 10, pp. 75 95 (1989). [11],, :,, 5, pp. 113 127, (1990). [12], (Mar. 1991). [13], (May, 1992) in [35]. [14] N. Otsu: Toward Flexible Information Processing: Theory and Novel Functions, Japan Computer Quarterly, JIPDEC, No. 89, pp. 9 17 (1992). [15] :,, Computer Today 5 No.49 (1992). [16] : RWC 92,, Vol. 25, No. 6 (1992). [17],, 4 (1993). [18] N. Otsu: Toward Flexible Intelligence MITI s New Program of Real World Computing, invited paper, Proc. IJCAI-93, Vol. 1, pp. 786 791 (Chambery, 1993). [19] N. Otsu: Real World Computing Program Overview, Theory and Novel Functions, Proc. IJCNN-93, pp. 1065 1070 (Nagoya, 1993). [20] : RWC,, Vol. 12 (1993). [21] :,, RWC, Vol. 34, No. 12 (1993). [22] :,, Vol. 9, No. 5, pp. 358 364 (1994). [23] Zadeh: Fuzzy Logic, Neural Networks, and Soft Computing, Communication of the ACM, Vol. 37, No. 3, pp. 77 84 (March 1994). [24] : RWC, RWC NEWS, Vol. 1 (Apr. 1995). [25] : RWC, RWC NEWS, Vol. 2 (Jul. 1995). [26] RWC 1997 (Jun. 1997), in RWC NEWS, Vol. 8 (May. 1997). [27], RWC 1997, ibid (1997). 25
[28] :, RWC 1997, ibid (1997). [29] :, RWC 1997, ibid (1997). [30] :, RWC 1997, ibid (1997). [31] (RWC-RWI/PDC), (May, 1997) in [35]. [32] (RWC-RWI/PDC),, RWC NEWS, Vol. 9 (Aug. 1997). [33],, : RWC, RWC NEWS, Vol. 10 (Nov. 1997). [34] RWC 2001 (Oct. 2001), in RWC NEWS, Vol. 20 (Nov. 2001). [35], (Feb. 2002) at [1]. [36], WG (Feb. 2002) at [1]. [37] :,, Vol. 122, No. 4, pp. 240-243 (2002). [38] : RWC,, Vol. 17, No. 2 (2002). [39] :, (https://staff.aist.go.jp/y.motomura/bn2002/paper/otsu.pdf) (2002). [40],, 292, (2004). A: 1970 f(r) C j y i n n y Y p(y C j ) P (C j ) (y) P (C j y) =P (C j )p(y C j )/p(y) Chow, 1957 f 26
f x x = Φ[f] T (λ) Φ inv [T (λ)f] =Φ inv [f] Lie ξ = g(r)f(r) dr g ξ x i λ λ =Φ var [T (λ)f] Φ var [f] f T (λ) [4][6] T (λ)f X x x m X n <m Y y = Ψ(x) Ψ Y X Ψ m n A y = A x A A Y Y B Y, W Y Ψ J[Ψ] = tr (W 1 Y B Y ) X B X, W X A B X a i = λ i W X a i n K n min(k 1,m) Fisher Fisher, 1936 2 y =Ψ D(x) = K P (C j x)c j j=1 27
Y c j X K K S =[s ij ], s ij = P (C j x)p(x C i ) dx = P (C j C i ) n n K 1 [5][6] p(x C j ) Y K 1 ε 2 [Ψ] = K j=1 P (C j ) Ψ(x) e j 2 p(x C j ) dx Y e j (j 1 e i e j = δ ij ) C j δε 2 [Ψ] = 2 K j=1 P (C j ){Ψ(x) e j }p(x C j ) δψ(x) dx =0 δψ(x) [6][7] K y =Ψ P (C j )p(x C j ) K R(x) = e j = P (C j x) e j p(x) j=1 e j ε 2 [Ψ] 1986 Ψ R (x) ε 2 [Ψ R]=1 tr Γ Γ =[γ ij ] K K γ ij = P (C i x)p (C j x)p(x)dx = P (C i C j ) s ij j=1 P (C i )s ij = P (C i )P (C j C i )=P (C i C j )=γ ij S Γ P (C i C i ) 0 D M E[M] = D M 2 + kρ(m) min for M ρ(m) k 28
e E[M] = e D M 2 e kρ(m) max for M p(d M) p(m) e E[M] = p(d M)p(M) =p(m D)p(D) max for M D p(m D) M B: [36] RWI 19) RWCP RWCP RWI. 19) RWI 2001 29
GA RHiNET SCore MPEG7 [36] ( ) 1971 3 4 1985 4 1990 4 1991 4 RWC (1992 2001 ) 1997 9 2001 4 1992 4 2010 3 2001 4 2007 3 2012 3 30