2d:null 2e:["$","title",null,{"children":"짧은 머신러닝 안내서:RANSAC(RANdom SAmple Consensus) | Humble Data Miner | Observable"}] 33:T6f4,{ const ps = 300; // square plot size const margin = { left:10, right:10, top:10, bottom: 10 }; const svg = d3.create("svg").attr("viewBox", [0, 0, ps, ps]).attr("width",ps).attr("height",ps); const urng = d3.randomUniform(); const nrng = d3.randomNormal(); const sub1 = d3.range(100).map(_=>{ const x=urng()*0.7; return {x:x, y:0.2*x+nrng()/200}; }); const sub2 = d3.range(30).map(_=>{ const x=urng()*0.2+0.7; return {x:x, y:0.5*x-0.23+ nrng()/200}; }); const sub3 = d3.range(30).map(_=>{ const x=urng()*0.1+0.9; return {x:x, y:1.2*x-0.88+ nrng()/200}; }); const sub = sub1.concat(sub2).concat(sub3); const ex = d3.extent(sub,d=>d.x); const ey = d3.extent(sub,d=>d.y); const x = d3.scaleLinear().domain(ex).range([margin.left, ps-margin.right]); const y = d3.scaleLinear().domain(ey).range([ps-margin.bottom, margin.top]); const color = d3.schemeCategory10[2]; const r = 2; const xGrid = (g) => g .attr('class', 'x-grid-lines') .selectAll('line') .data(x.ticks()) .join('line') .attr('stroke','gray').attr("stroke-opacity",0.2) .attr('x1', d => x(d)).attr('x2', d => x(d)) .attr('y1', margin.top).attr('y2', ps - margin.bottom); svg.append("g").call(xGrid); const yGrid = (g) => g .attr('class', 'y-grid-lines') .selectAll('line') .data(y.ticks()) .join('line') .attr('stroke','gray').attr("stroke-opacity",0.2) .attr('x1', margin.left) .attr('x2', width - margin.right) .attr('y1', d => y(d)) .attr('y2', d => y(d)) svg.append("g").call(yGrid); svg.append("g").attr("stroke","gray").attr("stroke-opacity",0.5).attr("fill",color).selectAll("circle") .data(sub) .join("circle") .attr("r",r) .attr("cx",d=>x(d.x)) .attr("cy",d=>y(d.y)); return svg.node(); }34:T601,--- ## 알고리즘 개요 * 매 반복(iteration)마다 후보모델을 하나 만들고 최적모델과 비교후 채택/기각 1. 지정된 랜덤샘플수(min_samples)만큼의 표본을 랜덤 결정 2. 랜덤표본들만을 기준으로 랜덤표본모델 생성 3. 랜덤표본모델을 기준으로 ${textcolor("경계거리(threshold)내",color_map.band)}의 점들을 ${textcolor("inlier",color_map.inlier)}로 규정 4. 현재의 ${textcolor("inlier모델",color_map.inlier)}을 ${textcolor("직전최적모델",color_map.best)}과 비교해 채택 또는 기각 * 비교기준: inlier수가 더 많은 모델 또는 에러(e.g RMSE)가 적은 모델 ## 알고리즘 동작 예시 * 데이터셋: [최신 NVIDIA 그래픽카드 이슈털기](https://observablehq.com/@humble-data-miner/looking-into-issues-of-nvidia-graphic-cards-esp-for-40-series)의 30시리즈 CUDA-MSRP 데이터 * 파라메터 변경 또는 '다시계산 '버튼으로 최적모델 재계산 가능 * 더 나은 모델 선정기준: inlier수가 더 많은 모델 (동률일때는 에러가 적은 모델) * 반복회차를 늘려가면서 알고리즘의 순차적 진행을 애니메이션으로 시각화35:T659,viewof ransac_params = { const dim = {w:340*2,h:75}; const svg = d3.create("svg") .attr("viewBox", [0, 0, dim.w, dim.h]) .attr("width", dim.w) .attr("height", dim.h); const samples = d3.range(2,6); const slider_samples = d3_slider.sliderBottom().min(d3.min(samples)).max(d3.max(samples)) .tickValues(samples).tickFormat(v=>v.toFixed(0)) .value(samples[1]).step(1).width(280).fill(d3.schemeCategory10[1]).on("onchange", () => svg.dispatch("input")); const gsamples = svg.append("g").attr("class","ransac-param-samples").attr("transform",`translate(20,40)`); gsamples.call(slider_samples); gsamples.append("text") .attr("text-anchor","middle").attr("dominant-baseline","hanging") .attr("x","140").attr("y",-30).attr("dy",4).text("랜덤샘플수(min_samples)"); const thresholds = d3.range(50,550,50); const slider_threshold = d3_slider.sliderBottom().min(d3.min(thresholds)).max(d3.max(thresholds)).tickValues(thresholds) .value(thresholds[1]).step(50).width(280).fill(d3.schemeCategory10[0]).on("onchange", () => svg.dispatch("input")); const gthreshold = svg.append("g").attr("class","ransac-param-threshold").attr("transform",`translate(${dim.w/2+20},40)`); gthreshold.call(slider_threshold); gthreshold.append("text") .attr("text-anchor","middle").attr("dominant-baseline","hanging") .attr("x","140").attr("y",-30).attr("dy",4).text("포함배제-경계거리(residue-threshold)"); return Object.defineProperty( svg.node(), "value", {get: () => ({min_samples:slider_samples.value(), threshold:slider_threshold.value()})} ); }36:T2473,{ const height = 500; const slider_height = 60; const margin = { left: 40, right:40, top:0, bottom: 20+slider_height }; const svg = d3.create("svg").attr("viewBox", [0, 0, width, height]); const data = rawdata; const ex = [d3.min(data,d=>d.CUDA)-1000,d3.max(data,d=>d.CUDA)+1000]; const x = d3.scaleLinear().domain(ex).range([margin.left, width-margin.right]); const y = d3.scaleLinear().domain([0,d3.max(data,d=>d.MSRP)+100]).range([height-margin.bottom, margin.top]); const color = d3.schemeCategory10[2]; const dot_radius = 5; const legend = svg.append("g"); legend.append("circle") .attr("stroke",d3.schemeCategory10[1]).attr("stroke-dasharray","6 6").attr("stroke-width",6).attr("stroke-opacity",0.7).attr("fill","none") .attr("cx", margin.left+30).attr("cy", margin.top+50).attr("r", dot_radius+10); legend.append("text") .attr("text-anchor","start").attr("dominant-baseline","middle").attr("x", margin.left+50).attr("y", margin.top+50).text("랜덤샘플"); legend.append("circle") .attr("fill",color).attr("r",dot_radius+10).attr("cx",margin.left+30).attr("cy",margin.top+90); const cur_inlier = legend.append("text") .attr("fill","white").attr("text-anchor","middle").attr("dominant-baseline","middle").attr("font-weight","bold") .attr("x", margin.left+30).attr("y", margin.top+90).attr("dy",2).text(""); const cur_model = legend.append("text") .attr("text-anchor","start").attr("dominant-baseline","middle") .attr("x", margin.left+50).attr("y", margin.top+90).text("현재모델: "); legend.append("circle") .attr("fill","grey").attr("r",dot_radius+10).attr("cx",margin.left+30).attr("cy",margin.top+130); const best_inlier = legend.append("text") .attr("fill","white").attr("text-anchor","middle").attr("dominant-baseline","middle").attr("font-weight","bold") .attr("x", margin.left+30).attr("y", margin.top+130).attr("dy",2).text(""); const best_model = legend.append("text") .attr("text-anchor","start").attr("dominant-baseline","middle") .attr("x", margin.left+50).attr("y", margin.top+130).text("직전최적모델: "); const sels = svg.append("g").attr("fill","none").selectAll("circle") .data(data) .join("circle") .attr("stroke",d3.schemeCategory10[1]).attr("stroke-dasharray","6 6").attr("stroke-width",6).attr("stroke-opacity",0) .attr("cx", d=>x(d.CUDA)).attr("cy", d=>y(d.MSRP)).attr("r", dot_radius+10); const dots = svg.append("g").selectAll("circle") .data(data) .join("circle") .attr("stroke",color).attr("stroke-width",2).attr("fill",color).attr("fill-opacity",0) .attr("cx", d=>x(d.CUDA)).attr("cy", d=>y(d.MSRP)).attr("r", dot_radius); const labels = svg.append("g") .attr("font-size",10).attr("text-anchor","start").attr("dominant-baseline","hanging").selectAll("text") .data(data) .join("text") .attr("x", d=>x(d.CUDA)).attr("y", d=>y(d.MSRP)) .attr("dx",dot_radius).attr("dy",dot_radius) .text(d=>d.Abbrv); function gen_line(model){ const line = d3.line().x(d=>x(d.x)).y(d=>y(d.y)); return line([{x:ex[0],y:model[0]+model[1]*ex[0]},{x:ex[1],y:model[0]+model[1]*ex[1]}]); } function gen_band(model){ const bw = ransac_params.threshold; const line = d3.line().x(d=>x(d.x)).y(d=>y(d.y)); return line([{x:ex[0],y:model[0]+model[1]*ex[0]+bw}, {x:ex[1],y:model[0]+model[1]*ex[1]+bw}, {x:ex[1],y:model[0]+model[1]*ex[1]-bw}, {x:ex[0],y:model[0]+model[1]*ex[0]-bw}]); } const band_sel = svg.append("path").attr("stroke","none").attr("fill",d3.schemeCategory10[0]).attr("opacity",0).attr("d",gen_band(ransac_result.all.model)); const line_sel = svg.append("path").attr("stroke",d3.schemeCategory10[0]).attr("stroke-width",2) .attr("stroke-dasharray","2 2").attr("fill","none").attr("opacity",0) .attr("d",gen_line(ransac_result.all.model)); const line_best = svg.append("path").attr("stroke",d3.schemeCategory10[2]).attr("stroke-width",2) .attr("fill","none").attr("opacity",0) .attr("d",""); svg.append("path").attr("fill","white").attr("stroke","none").attr("d",`M 0 0 H ${margin.left} V ${height} H 0 V 0`); svg.append("path").attr("fill","white").attr("stroke","none").attr("d",`M 0 ${height-margin.bottom} H ${width} V ${height} H 0`); svg.append("g").append("text") .attr("text-anchor","end").attr("dominant-baseline","auto").attr("dx",-4).attr("dy",-4) .attr("x", width-margin.right).attr("y",height-margin.bottom).text("x: CUDA 코어수"); svg.append("g").append("text") .attr("text-anchor","start").attr("dominant-baseline","hanging").attr("dx",4).attr("dy",4) .attr("x", margin.left).attr("y",margin.top).text("y: MSRP(제조사 권장 소비자 가격)"); yield svg.node(); const history = ransac_result.history; const its = d3.range(-1,history.length,1); const slider = d3_slider.sliderBottom().min(-1).max(its.length-2).step(1).value(-1) .tickValues(its) .tickFormat(i=>`${(i+1).toFixed(0)}${i<0?"":history[Math.round(i)].updated?'(채택)':"(기각)"}`) .value(-1).width(width-160).fill(d3.schemeCategory10[3]).on("onchange", () => update(slider.value())); const giteration = svg.append("g").attr("class","ransac-iteration").attr("transform",`translate(100,${height-margin.bottom/2})`); giteration.call(slider); giteration.selectAll(".slider text").attr("font-size", "12"); giteration.selectAll(".axis text").attr("fill",i=>i>=0&&history[i].updated?"green":"grey"); giteration.append("text") .attr("text-anchor","middle").attr("dominant-baseline","middle") .attr("x","-50").attr("y",0).attr("dy",4).text("반복회차"); giteration.append("text") .attr("text-anchor","middle").attr("dominant-baseline","middle") .attr("x","-50").attr("y",20).attr("dy",4).text("(#iteration)"); svg.append("g").attr("transform",`translate(0,${height-margin.bottom})`).call(d3.axisBottom(x)); svg.append("g").attr("transform",`translate(${margin.left},0)`).call(d3.axisLeft(y).tickFormat(d=>`$${d.toLocaleString()}`)); function update(it) { it = Math.round(it); const dur = anim_step_ms; if(it<0) { giteration.selectAll(".slider text").attr("fill","grey").attr("font-weight","bold"); band_sel.transition().duration(dur).attr("opacity",0); line_sel.transition().duration(dur).attr("opacity",0); line_best.transition().duration(dur).attr("opacity",0); sels.transition().duration(dur).attr("stroke-opacity",0); dots.transition().duration(dur).attr("fill-opacity",0).attr("stroke",color); cur_inlier.text(""); cur_model.text("현재모델:"); best_inlier.text(""); best_model.text("직전최적모델:"); return; } const hitem = history[it]; const prevbest = it>0?history[it-1].best:null; giteration.selectAll(".slider text").attr("fill",hitem.updated?"green":"grey").attr("font-weight","bold"); cur_inlier.text("").transition().delay(dur*5).text(hitem.ids_in.reduce((acc,cur)=>{acc+=cur?1:0; return acc},0)); const cur_model_text = "현재모델: "+ (!hitem.candidate? "(inlier가 부족해서 기각됨)": `y=${hitem.candidate.model[1].toFixed(3)}x${hitem.candidate.model[0]>=0?"+":""}${hitem.candidate.model[0].toFixed(0)} (RMSE:${hitem.candidate.rmse.toFixed(2)})`); cur_model.text("현재모델:").transition().delay(dur*5).text(cur_model_text); best_inlier.text(prevbest?prevbest.inliers.length:""); best_model.text(prevbest?`직전최적모델: y=${prevbest.model[1].toFixed(3)}x${prevbest.model[0]>=0?"+":""}${prevbest.model[0].toFixed(0)} (RMSE:${prevbest.rmse.toFixed(2)})`:"직전최적모델:"); sels. transition().duration(dur/2).attr("stroke-opacity",0). transition().duration(dur/2).attr("stroke-opacity",(d,i)=>hitem.ids_sel[i]?0.7:0). transition().duration(dur/2).attr("stroke-opacity",0). transition().duration(dur/2).attr("stroke-opacity",(d,i)=>hitem.ids_sel[i]?0.7:0); line_sel. transition().duration(dur/2).attr("opacity",0). transition().duration(0).attr("d",gen_line(hitem.model_sel)). transition().delay(dur*3/2).duration(dur).attr("opacity",1); band_sel. transition().duration(dur/2).attr("opacity",0). transition().duration(0).attr("d",gen_band(hitem.model_sel)). transition().delay(dur*5/2).duration(dur).attr("opacity",0.2); dots.transition().duration(dur/2).attr("fill-opacity",0). transition().delay(dur*7/2).duration(dur/2).attr("fill-opacity",(d,i)=>hitem.ids_in[i]?1:0). transition().duration(dur/2).attr("fill-opacity",0). transition().duration(dur/2).attr("fill-opacity",(d,i)=>hitem.ids_in[i]?1:0); dots.attr("stroke",(_,i)=>prevbest && prevbest.ids_in[i]?"grey":color); line_best.attr("stroke","black").attr("opacity",0.5).attr("d",prevbest?gen_line(prevbest.model):""); if(hitem.best) { if(hitem.updated) { if(!prevbest) line_best.attr("stroke",color).attr("opacity",0); line_best. transition().delay(dur*10/2).duration(dur).attr("opacity",0.5).attr("d",gen_line(hitem.best.model)).attr("stroke",color); } } else line_best.transition().duration(dur/2).attr("opacity",0); } return svg.node(); }37:Tb1f,ransac_result = { rerun; const data = rawdata; const n = data.length; const X = mathjs.reshape(mathjs.matrix(data.map(d=>d.CUDA)),[n,-1]); const y = mathjs.reshape(mathjs.matrix(data.map(d=>d.MSRP)),[n,-1]); function fit(X,y) { const aX = mathjs.concat(mathjs.ones(mathjs.size(X).get([0]),1),X); const fit_res = mathjs.multiply( mathjs.multiply( mathjs.inv(mathjs.multiply(mathjs.transpose(aX),aX)), mathjs.transpose(aX) ), y ); return fit_res; } function predict(model,X) { const aX = mathjs.concat(mathjs.ones(mathjs.size(X).get([0]),1),X); return mathjs.multiply(aX,model); } function loss(model,X,y) { const pred = predict(model,X); return mathjs.sqrt(mathjs.sum(mathjs.map(mathjs.subtract(pred,y),mathjs.square))/mathjs.size(X).get([0])); } const history = []; const params = {n:ransac_params.min_samples,k:20,t:ransac_params.threshold}; const ids = d3.range(n); let best = {inliers:[], rmse:Infinity, model:null}; for(let it = 0; it < params.k; it++) { const hitem = {}; // model fit to sampled subset of length params.n const ids_sel = d3.shuffle(ids).slice(0,params.n); const set_sel = new Set(ids_sel); const mask_sel = mathjs.index(ids_sel,0); const model_sel = fit(mathjs.subset(X,mask_sel), mathjs.subset(y,mask_sel)); hitem.ids_sel = ids.map((_,i)=>set_sel.has(i)?true:false); hitem.model_sel = Array.from(model_sel).map(v=>v.value); // all datapoints prediction to the subset model const pred_all = predict(model_sel, X); const diff_cur = Array.from(mathjs.flatten(mathjs.abs(mathjs.subtract(y,pred_all)))); const ids_cur = diff_cur.filter(v=>v.value<=params.t).map(v=>v.index[0]); const set_cur = new Set(ids_cur); hitem.ids_in = ids.map((_,i)=>set_cur.has(i)?true:false); hitem.best = best.model ? best : null; hitem.updated = false; hitem.candidate = null; // continue when not enough inliers or less inliers than the best ever if(ids_cur.length < 2 || ids_cur.length < best.inliers.length) { history.push(hitem); continue; } const mask_cur = mathjs.index(ids_cur,0); const Xsub = mathjs.subset(X,mask_cur); const ysub = mathjs.subset(y,mask_cur); const model_sub = fit(Xsub,ysub); const loss_cur = loss(model_sub, Xsub, ysub); hitem.candidate = {model:Array.from(model_sub).map(v=>v.value), rmse: loss_cur}; if(best.inliers.length == ids_cur.length && best.rmse <= loss_cur) { history.push(hitem); continue; } best = {inliers: ids_cur, ids_in:hitem.ids_in, rmse: loss_cur, model: Array.from(model_sub).map(v=>v.value)}; hitem.best = best; hitem.updated = true; history.push(hitem); } return {all:{model:Array.from(fit(X,y)).map(v=>v.value)}, history: history} }38:[{"id":"9b7467ed1b6547add478b614706835649e4e1e6732e4a41cb1cd8641b4019c7eaea2ccd8db7bc61b27ee9ceb89f14c4004567ad6c5168f5d646242c43e805c32","href":"https://static.observableusercontent.com/files/9b7467ed1b6547add478b614706835649e4e1e6732e4a41cb1cd8641b4019c7eaea2ccd8db7bc61b27ee9ceb89f14c4004567ad6c5168f5d646242c43e805c32","name":"dataset(10,16,20,30,40).csv","type":"text/csv","lastModified":1692375966554.033,"size":3415}] 31:{"id":"0b0977d9022a326a","slug":"short-guide-to-ml-ransac","title":"짧은 머신러닝 안내서:RANSAC(RANdom SAmple Consensus)","thumbnail":"19a9b4c73cc870b2edb20b2ae492878c42fc917652b86ecda559d5d0eaf05778","access_level":"public","create_time":"$D2023-08-18T16:20:57.762Z","update_time":"$D2023-09-05T17:09:52.411Z","view_time":null,"first_public_time":"$D2023-08-18T16:20:57.762Z","has_unpublished_edits":null,"fork":false,"stars":0,"starred":false,"owner":{"type":"team","id":"50faa4da52a9ff9b","login":"humble-data-miner","name":"Humble Data Miner","avatar_url":"https://avatars-next.observableusercontent.com/3046b1bea86c5e825baea7c0c403d90ad9690b5396b11b5d02b3a0839fbcb946","tier":"starter_2024"},"version":888,"publish_version":null,"role":null,"authored":false,"authors":[{"type":"individual","id":"4e2fc41e06361cc3","login":"humble-data-miner","name":"Humble Data Miner","avatar_url":"https://avatars-next.observableusercontent.com/3046b1bea86c5e825baea7c0c403d90ad9690b5396b11b5d02b3a0839fbcb946","tier":null}],"members":[{"type":"individual","id":"4e2fc41e06361cc3","login":"humble-data-miner","name":"Humble Data Miner","avatar_url":"https://avatars-next.observableusercontent.com/3046b1bea86c5e825baea7c0c403d90ad9690b5396b11b5d02b3a0839fbcb946","tier":null,"role":"editor"}],"comments":[],"custom_thumbnail":true,"trashed":false,"lapsed":null,"subscribed":null,"forks":1,"collection":{"id":"b886df04acb387aa","slug":"short-guide-to-machine-learning","title":"Short Guide to Machine Learning","owner":{"id":"50faa4da52a9ff9b","login":"humble-data-miner","name":"Humble Data Miner"}},"type":"notebook","body":{"title":"짧은 머신러닝 안내서:RANSAC(RANdom SAmple Consensus)","cells":[{"id":0,"value":"# 짧은 머신러닝 안내서:
RANSAC(RANdom SAmple Consensus)","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":724,"value":"---\n## Random Sample Consensus?\n* 무작위(Random)로 일부 데이터를 뽑아(Sample) 단일 모델을 추정하고, 이를 반복해 그중 가장 합의(Consensus)된 모델을 찾는 알고리즘\n * Random Sample: 추정하려는 모델이 잘 들어맞는 부분 데이터들을 찾기 위함 → 들어맞는 데이터들(inliers), 맞지않는 데이터들(outliers)\n * Consensus(가장 합의된 모델): 부분데이터를 잘 설명하는 모델들중 가장 그럴듯한 모델 ⇔ 가장 많은 데이터포인트 혹은 가장 에러가 낮은 모델\n\n## 손으로 푸는 RANSAC\n* 아래 산포도를 대표하는 가장 유력한 직선 하나를 찾는다면?\n 1. 2개의 데이터포인트를 임의로 선택 (Random Sample)\n 2. 선택된 2개의 데이터포인트로 직선(선형모델) 그리기\n 3. 직선(선형모델)을 기준으로 인접 데이터포인트들(inlier)을 묶어 직선합의체(Consensus) 형성\n 4. 위 1,2,3 과정을 반복해 나온 결과중, 가장 데이터포인트들이 많이 소속된 직선합의체(Consensus)를 선택 → 대표직선(선형모델)","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":741,"value":"$33","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":35,"value":"$34","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":332,"value":"viewof rerun = Inputs.button(\"다시계산\")","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":216,"value":"$35","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":43,"value":"$36","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":687,"value":"## 알고리즘 결과로 보는 참고사항\n\n* 랜덤샘플수는 결과에 주는 영향이 작다\n * 위 예에서 min_samples=2로 충분 (#독립변수+1)\n* 경계거리값이 결과에 주는 영향이 크다 → 미세 조정이 필요\n * 위 예에서 threshold=500으로 설정시 모든점이 inlier가 되는 문제 발생\n* 모델의 선정기준에 따라 결과에 중요한 파라메터가 다름\n * inlier수가 많은 모델 선택: 경계거리값 선정이 중요\n * 에러가 적은 모델 선택: 최소 inlier수라는 추가 파라메터 필요\n * inlier수가 작을때 에러도 작은 경향 → 국소적 최적모델 선택가능\n \n## 알고리즘의 일반적인 활용\n * 유력한 단일모델을 찾는데 활용\n * 주로 선형모델에 쓰나 비선형모델(다항,지수 등등)에도 활용가능\n * 유력한 단일모델에서 제외된 점들을 이상치(outlier)로 탐지","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":710,"value":"---\n\n## 참고자료\n* 종합적인 RANSAC 설명자료: [Wikipedia - Random sample consensus](https://en.wikipedia.org/wiki/Random_sample_consensus)\n* Python RANSAC 모듈: [sklearn.linear_model.RANSACRegressor](https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.RANSACRegressor.html)\n* NVIDIA 지포스 30 시리즈 사양: [Wikipedia - GeForce 30 series](https://en.wikipedia.org/wiki/GeForce_30_series)","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":885,"value":"---\n## 부연설명 동영상\n\n","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":12,"value":"---\n## Appendix","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":27,"value":"rawdata = await FileAttachment(\"dataset(10,16,20,30,40).csv\").csv({typed: \"auto\"})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"rawdata","database":"$undefined"},{"id":80,"value":"$37","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":441,"value":"viewof anim_step_ms = Inputs.range([500,2000], {step:500,value:1000,label:\"애니매이션 단위시간(ms)\"});","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":665,"value":"color_map = {\n const band_color = d3.color(d3.schemeCategory10[0]);\n band_color.opacity=0.5;\n return {\n inlier:d3.schemeCategory10[2],\n candidate:d3.schemeCategory10[0],\n best:\"grey\",\n sample:d3.schemeCategory10[1],\n band:band_color.formatHex8()\n }\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":630,"value":"function textcolor(content, style = {}) {\n function yiq(color) {\n const {r, g, b} = d3.rgb(color);\n return (r * 299 + g * 587 + b * 114) / 1000 / 255; // returns values between 0 and 1\n }\n const {\n background,\n color = yiq(background) >= 0.7 ? \"#111\" : \"white\", \n padding = \"0 2px\",\n borderRadius = \"4px\",\n fontWeight = 500,\n ...rest\n } = typeof style === \"string\" ? {background: style} : style;\n return htl.html`${content}`;\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":14,"value":"html``","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":69,"value":"mathjs = require('mathjs');","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":220,"value":"d3_slider= require(\"d3-simple-slider\")","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"}],"files":"$W38","stdlib":"1","titling":"auto"}} 32:"$31:body"