# Fileset

[jjapae1d84supp1.pdf](https://mdr.nims.go.jp/filesets/420561ba-577d-4808-a42e-ae3ec54d8310/download)

## Creator

[Daiki Nishioka](https://orcid.org/0000-0002-3369-7700), Kaoru Shibata, [Wataru Namiki](https://orcid.org/0000-0003-4053-7366), [Kazuya Terabe](https://orcid.org/0000-0003-3988-3456), [Takashi Tsuchiya](https://orcid.org/0000-0002-6950-6160)

## Rights

[Creative Commons BY Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/)

## Other metadata

[Physical masking-induced enhancement of information processing capacity in a redox-type ion-gating reservoir](https://mdr.nims.go.jp/datasets/06ce0dfc-0c9c-4ec5-b4df-f582664d7ffb)

## Fulltext

LCO-IGR-PM_SI_REV1 1 Supplementary information Physical Masking-Induced Enhancement of Information Processing Capacity in a Redox-Type Ion-Gating Reservoir  Daiki Nishioka1*, Kaoru Shibata2,3, Wataru Namiki2, Kazuya Terabe2, Takashi Tsuchiya2,3*  1International Center for Young Scientists (ICYS), National Institute for Materials Science (NIMS), 1-1 Namiki, Tsukuba, Ibaraki, 305-0044, Japan.  2Research Center for Materials Nanoarchitectonics (MANA), National Institute for Materials Science (NIMS), 1-1 Namiki, Tsukuba, Ibaraki, 305-0044, Japan.  3Department of Applied Physics, Tokyo University of Science, Katsushika, Tokyo 125-8585, Japan   *Corresponding author NISHIOKA.Daiki@nims.go.jp, TSUCHIYA.Takashi@nims.go.jp  2  Supplementary Fig. 1 | Effect of physical masking conditions on the computational performance of the redox-type IGR. Normalized mean squared error (NMSE, top) in the NARMA2 task and information processing capacity (IPC, bottom) as functions of the amplitude (VD1,2) and frequency (fD1,2) of the triangular drain voltages. To focus solely on the effect of the physical masking conditions, the results shown here were obtained without using the inveted input method and were based only on 60 reservoir states.  1614121086420IPC0.120.100.080.060.040.02NMSEVD1,2=±400 mVVD1,2=±800 mVf D1,2=100 mHzf D1,2=100 mHzf D1,2=250 mHzf D1,2=250 mHzTrainTestMixed conditionVD1=±400 mVVD2=±800 mVfD1 = 100 mHzfD2 = 250 mHz100IPC      1st order100IPC      2nd order3rd order4th order5th order6th order 3  Supplementary Fig. 2 | Gate voltage VG input applied to the device and the corresponding current responses (left). The right panel shows the virtual nodes sampled at discrete time step k, indicated by circular markers.  -2-10VG /V-20-1001020I G /nA-404I D1 /µA-202I D2 /µA200180160140120100Time /s-20-1001020I G /nA-404I D1 /µA-202I D2 /µA190180170Time /s-2-10VG /VInputIGID1ID2Time step k k+1k–1x1x2x6x20………x21x23x40x39… …x41x42x60……1 s 4  Supplementary Fig. 3 | Schematic illustration of the six-fold cross-validation procedure used for the NARMA2 task. Among the total of 1250 data points, the initial 50 data points were used for reservoir washout, and the remaining 1200 data points were subjected to six-fold cross-validation. The results shown in the main text correspond to those obtained from the bottom segment of the schematic.    Supplementary Table 1 | Results of the six-fold cross-validation for the NARMA2 task.  NMSE Train range Test range Train Test 250<k≤1250 50<k≤250 0.0219 0.0442 50<k≤250, 450<k≤1250 250<k≤450 0.0209 0.0316 50<k≤450, 650<k≤1250 450<k≤650 0.0211 0.0322 50<k≤650, 850<k≤1250 650<k≤850 0.0205 0.0357 50<k≤850, 1050<k≤1250 850<k≤1050 0.0216 0.0270 0<k≤1050 1050<k≤1250 0.0215 0.0333 Average 0.0212 0.0340  200 steps 1000 stepsWashout:50 stepsTime step k… 6-fold cross validationTesting Training