3f:null 40:["$","title",null,{"children":"Continuous Time Networks (CTRNN) | Dan Brickley | Observable"}] 46:T4b5,// This is the strangest example, skip it unless working on alternate activation functions function osc2NodeExampleAlternate() { let a = 0.5; // 0 is sin, 1 is tanh. a = a_param; let net = ctNet({ size: 2, // squashfn: (x) => tf.sigmoid(x) // <- default so not needed. //squashfn: (x) => tf.sin(x), // instead of logistic sigmoid squashfn: (x) => tf.sin(x), // not interesting but this is where you'd do sin+tanx ... // couldn't get this working, maybe due to vectorized nature of the function we need // // // a* and 1-a * ... /* see https://edizquierdo.wordpress.com/2006/12/09/generative-music-creative-dynamical-systems-and-ctrnn-activation-functions/ Where F=A*tanh + (1-A)*sin, thus A=1 is the simple tanh, A=0 is the sine wave, A=0.5 is what we just showed. As soon as we get closer to the tanh (A=0.75) the system begins to get more ‘classical’ looking... */ init_weights: [ [4.5, 1], [-1, 4.5] ] }); net.states = tf.tensor1d([2, 3]); net.biases = tf.tensor1d([-2.75, -1.75]); // net.run_duration = 100000; net.step_size = 1; net.name = "2-Node, Alt activation function."; Object.freeze(net); return net; }47:T1328,// τy' = -y + Wσ(g(y+ θ)) + I function ctNet(spec) { if (spec === undefined) { spec = {}; } let override_squashing_function = spec["squashfn"] || false; if (typeof spec === "number") { spec = { size: spec }; } // avoiding use of this., clumsily. function docalc(net, livestates, liveoutputs) { // todo: add input_weights? /* yprime = -states + total_inputs delta = yprime * step_size / tau states += delta outputs = tf.sigmoid(gain*(states+bias)) */ // activation function // vectorized. function squash(x) { // return tf.tanh(x); // return tf.sigmoid(x); if (override_squashing_function == false) { return tf.sigmoid(x); } else { // console.log("Alternate squashing function!", spec["squashfn"]); return spec["squashfn"](x); } } // todo: api for other functions // eg @ylecun // 1.7159 * tanh((2x/3) // http://yann.lecun.com/exdb/publis/pdf/lecun-98b.pdf // can we try to show sweep with these parqms? // more! // https://edizquierdo.wordpress.com/2006/12/09/generative-music-creative-dynamical-systems-and-ctrnn-activation-functions/ // eg "...half way between sin and tanh" const calc = tf.tidy(() => { let total_inputs = net.weights.dot(liveoutputs).add(net.ext_inputs); /* todo: we are leaking yprime and output tensors, hence cleanup downstream from mainloop. Address here? */ // let oldgradients = yprime; let yprime = total_inputs.sub(livestates); //.div(net.taus); // y' = -y +g(Wo+ϑ) //oldgradients.dispose(); // junked! let delta = yprime.mul(tf.scalar(net.step_size)).div(net.taus); // Euler step let trash = livestates; livestates = livestates.add(delta); trash.dispose(); liveoutputs = squash( livestates.transpose().add(net.biases).mul(net.gains) ); //σ // somewhere we need this - problem has been figuring out when to do it // new approach - copy to cpu each cycle. Expensive? should make it a flag and do some perf timings. Could also be done // every n loops... // coreloop.outputs.dispose(); //coreloop.yprime.dispose(); if (livestates.arraySync().includes(NaN)) { nanlog.push(livestates.arraySync()); throw { states: livestates }; } return { states: livestates, states_cpu: livestates.arraySync(), outputs: liveoutputs, outputs_cpu: liveoutputs.arraySync(), yprime: yprime, yprime_cpu: yprime.arraySync() }; }); //---- return calc; } /* Have the dynamics settled down? logic in js i.e. CPU for now TODO: use tf.whereAsync to use GPU */ function in_fixed_point(yp) { return tf.tidy(() => { const mask = yp.equal([0]).asType("bool"); const isz = mask.arraySync().every((item) => item !== 0); return isz; }); } const ones1d = (n) => { return tf.ones([n]); }; const genRand2d = (n) => tf.tensor2d( Array.from({ length: n * n }, () => Math.random() * 2), [size, size] ); const genFixed2d = (n) => tf.tensor2d(Array(size * size).fill(0.1), [size, size]); const zeros1d = (n) => { return tf.zeros([n]); }; if (!spec.size) { spec.size = 3; } // todo: move to a principled Examples structure let size = spec.size; var net = { size: spec.size }; net.ones1d = ones1d; net.zeros1d = zeros1d; if (spec.init_weights) { if (Array.isArray(spec.init_weights)) { // amon.log("init weights as array: ", spec.init_weights); net.weights = tf.tensor2d(spec.init_weights, [size, size]); } else { if (typeof spec.init_weights === "function") { let values = Array.from({ length: size * size }, spec.init_weights); net.weights = tf.tensor2d(values, [size, size]); } else { throw ( ("init_weights should be an array, a or a function.", spec.init_weights) ); } } } else { net.weights = genRand2d(size); } net.biases = zeros1d(size); net.taus = ones1d(size); net.gains = ones1d(size); // things that change (will be generators not mutables): net.ext_inputs = zeros1d(size); net.states = zeros1d(size); net.outputs = zeros1d(size); // todo - expose functions as an api here net.eulerStep = docalc; net.in_fixed_point = in_fixed_point; // amon.log("net created: ", net); net.run_duration = 10000; //net.freeze = function () {}; net.freeze = function (net, weights) { if (net.weights != undefined) { net.weights = weights; net.weights_cpu = net.weights.arraySync(); Object.freeze(net); } else { console.log("Can't freeze a weightless network."); // throw exception? throw "Can't freeze a network with no weight matrix."; } }; return net; }48:T510,function net2dot(net) { let w = coreloop.net.weights_cpu; // ${n0} -> ${n0} [color=blue penwidth=5]; function edgelist_for_viz(net) { let dotlist = []; function w2ink(x) { if (x < -2) { return 5; } if (x < -1) { return 3; } if (x < -0.5) { return 1; } if (x > 2) { return 5; } if (x > 1) { return 3; } if (x > 0.5) { return 1; } // yeah yeah this is vile } let edgecol = "blue"; for (var fromNode in w) { for (var toNode in w[fromNode]) { if (w[fromNode][toNode] > 0) { edgecol = "blue"; } else { edgecol = "red"; } dotlist.push( `${fromNode} -> ${toNode} [color=${edgecol} penwidth=${w2ink( w[fromNode][toNode] )}]\n` ); } } return dotlist.join("\n"); } let dotintro = `digraph { graph [fontname = helvetica]; node [fontname = helvetica]; edge [fontname = helvetica]; node [style=filled]`; // generate node info let dotnode = ``; for (var n in w) { dotnode = dotnode + `# node ${n} `; } let dotedges = edgelist_for_viz(coreloop.net); return dot`${dotintro.concat(dotedges.concat("}")).concat(dotnode)}`; }49:[] 44:{"id":"283a9ba787236674","slug":"ctnlib","title":"Continuous Time Networks (CTRNN)","thumbnail":"1cd8f14f41b15d6c58f0d271947edc803e166b6bcc48a91e65680b366fd180b2","access_level":"public","create_time":"$D2021-10-01T08:37:27.963Z","update_time":"$D2024-12-04T13:54:11.382Z","view_time":null,"first_public_time":"$D2021-10-01T18:32:26.728Z","has_unpublished_edits":null,"fork":false,"stars":3,"starred":false,"owner":{"type":"team","id":"6aa75e0ed4c47ade","login":"mazeballsold","name":"Dan Brickley","avatar_url":"https://avatars-next.observableusercontent.com/c1e798353c1834a30de7804f61bbfd5116503c44db4803547a43228e89afc8cb","tier":"starter_2024","bio":"broken account, please ignore - migrating","following":false},"version":4268,"publish_version":4268,"role":null,"authored":false,"authors":[{"type":"individual","id":"44d3c304df1c5e78","login":"mazeballsold","name":"Dan Brickley","avatar_url":"https://avatars-next.observableusercontent.com/c1e798353c1834a30de7804f61bbfd5116503c44db4803547a43228e89afc8cb","tier":null,"bio":"broken account, please ignore - migrating","following":false}],"members":[{"type":"individual","id":"44d3c304df1c5e78","login":"mazeballsold","name":"Dan Brickley","avatar_url":"https://avatars-next.observableusercontent.com/c1e798353c1834a30de7804f61bbfd5116503c44db4803547a43228e89afc8cb","tier":null,"role":"editor"}],"comments":[],"custom_thumbnail":false,"trashed":false,"lapsed":null,"subscribed":null,"forks":2,"collection":null,"type":"notebook","body":{"titling":"auto","title":"Continuous Time Networks (CTRNN)","cells":[{"id":0,"value":"# Continuous Time Networks (CTRNN)","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2130,"value":"@danbri's CTRNN work in progress. I am trying to distill this into a .js module, and breaking things in the process. GPU leaks - rising tensorCount means a .dispose() needed somewhere. Personal scratchpad, not for re-use / sharing at this time! i.e. unlicensed for now. Specifically, I don't have my employer permission to opensource this yet, so please do not use it for anything! ","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":4218,"value":"mutable nanlog = []","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":4259,"value":"print(testrand)","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":4179,"value":"ones1d = (n) => {\n return tf.ones([n]);\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":4177,"value":"ones1d(16).arraySync()","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3847,"value":"viewof chosen_example = Inputs.select(\n examples,\n\n {\n label: \"Examples:\",\n format: (x) => x.name,\n value: examples.find((t) => t.name === \"RandomNet\")\n // value: examples.find((t) => t.name === \"Simple 4-Node\")\n }\n)","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":1686,"value":"liveplot = {\n let net = coreloop;\n let win = maxresults;\n let myMarks = makePlotMarks(net);\n yield Plot.plot({\n x: {\n axis: null\n },\n marks: myMarks\n /* something like this, ... [\n Plot.lineY(col(results, 0).slice(-win), { stroke: \"steelblue\" }),\n Plot.lineY(col(results, 1).slice(-win), { stroke: \"orange\" }),\n Plot.lineY(col(results, 2).slice(-win), { stroke: \"green\" })\n ] */\n });\n}\n\n// consider plotting node outputs rather than state? stack both plots? needs better results[] structure.","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3987,"value":"md`# Results data\\n ${JSON.stringify(results)} `","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":4222,"value":"nanlog","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":4054,"value":"results.slice(-15)","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3531,"value":"viewof reset_button = Inputs.button(\"Reset\")","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":1692,"value":"viewof maxresults = Inputs.range([10, 1800], {\n value: 200,\n step: 10,\n label: \"window zoom\"\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3737,"value":"viewof a_param = Inputs.range([0, 1], {\n value: 0.5,\n step: 0.01,\n label: \"Parameter 'a' (these are unused by default)\"\n}) // a general purpose parameter we can use to mess with below","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3747,"value":"viewof b_param = Inputs.range([0, 1], {\n value: 0.3,\n step: 0.01,\n label: \"Parameter 'b'\"\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":1882,"value":"viewof logger = new Monitor()","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3949,"value":"coreloop.net.size","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":4074,"value":"coreloop.net.weights.arraySync()[0][1]","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3946,"value":"viewof scattercol_x = Inputs.range([0, 100], {\n value: 0,\n step: 1,\n label: \"scatterplot x-axis node\"\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3978,"value":"viewof scattercol_y = Inputs.range([0, 100], {\n value: 1,\n step: 1,\n label: \"scatterplot y-axis node\"\n})\n\n// hack but consider using -1 to represent PCA 1st dimension, https://observablehq.com/@jerdak/tensorflow-js-dimensionality-reduction-autoencoder ...? demo https://observablehq.com/d/0e21620e91961457 with CTRNN data.","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":1771,"value":"scatterplot = {\n slowloop; // visualize two dimensions/nodes only.\n var dataobjects = [];\n for (var row of results) {\n row = row.map((x) => x * 1000);\n var yp = coreloop.yprime_cpu;\n dataobjects.push({\n x: row[scattercol_x], // was: row[0],\n y: row[scattercol_y],\n z: row[2],\n total: row[0] + row[1] + row[2],\n yp: yp\n });\n // logger.log(\"xyz\", row);\n }\n\n yield Plot.plot({\n marks: [\n Plot.text(dataobjects, { x: \"X-Axis\", y: \"Y-Axis\", text: (d) => d.name }),\n Plot.dot(dataobjects, {\n x: \"x\", // { domain: [0 - 100] },\n y: \"y\",\n stroke: (d) => (d.in_fp ? \"red\" : \"black\") // todo\n })\n ]\n });\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3869,"value":"net2dot(slowloop.net)","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3883,"value":"import { table } from \"@binarmaya/table-from-array-of-data\" // proof-of-concept; better to make own, or use d3","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3885,"value":"table(slowloop.net.weights_cpu, {\n sortable: false,\n rank: false,\n header: false\n})","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":1980,"value":"mutable results = [] // this should be a more considered data structure, to put it mildly","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3817,"value":"chosen_example","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3858,"value":"chosen_example","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2896,"value":"coreloop = {\n //reset_button;\n //examples;\n // mutable results = [];\n return runSimulation(chosen_example);\n //return runSimulation(osc2NodeExampleAlternate());\n // return runSimulation(osc2NodeExample());\n // return = runSimulation(osc3NodeExample())\n // return runSimulation(osc4NodeExample());\n //return runSimulation(Random16Example());\n //return runSimulation(Beer1995Fig4a3a());\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3801,"value":"examples = [\n // RandomNetExample(),\n osc4NodeExample(),\n Beer1995Fig4a3a(),\n osc2NodeExample(),\n osc3NodeExample(),\n osc4NodeExample(),\n osc2NodeExampleAlternate(),\n Beer1995Example7()\n]","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3219,"value":"slowloop = {\n // sample slowly so we can do slower graphical ops like graphviz/dot.\n while (true) {\n if (coreloop.count % 10 == 0) {\n yield coreloop;\n }\n await Promises.delay(1);\n }\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3620,"value":"monitor_1 = {\n results.push(coreloop.outputs_cpu);\n\n logger.log(\n \"outputs\",\n coreloop.outputs_cpu.map((x) => ppfloat(x))\n );\n\n logger.log(\"numTensors\", tf.memory().numTensors);\n logger.log(\"count\", coreloop.count);\n logger.log(\"in a fixedpoint\", coreloop.in_fp);\n logger.log(\"size\", coreloop.net.size);\n logger.log(\"changey\", changey(coreloop));\n logger.log(\"run_duration\", coreloop.net.run_duration);\n logger.log(\"chosen_example\", chosen_example.name);\n //return runSimulation(chosen_example);\n\n // if i swap slowloop for coreloop on some of these, the whole block slows down. puzzling for now.\n\n // Take out the trash after delaying a second\n //\n await Promises.delay(1000);\n coreloop.outputs.dispose();\n //\"So the cause of the error you are seeing is that you are passing null values as inputs to the model. \" - for undefined reading backend bug.\n coreloop.yprime.dispose();\n // todo: push things into a trash buffer, and retire from the old end?\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2032,"value":"# Example Network definitions\n\nExamples including some that are variations of those from Randall Beer, On the Dynamics of Small Continuous-Time Recurrent Neural Networks* (1995). Specifically, from figure 3a:4a in Beer(1995):\n*this network is subtly different to 1st (also detailed in same paper) and does *not* oscillate).*\n- w 1 1: 5.25\n- w 1 2: -1\n- w 2 1: 1\n- w 2 2: 5.25\n- theta 1: -3.4\n- theta 2: -2.5\n\nNote that some properties should probably move to a simulation api. Need better setup for taus, biases too.","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3668,"value":"$46","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":974,"value":"function osc2NodeExample() {\n let net = ctNet({\n size: 2,\n init_weights: [\n [4.5, 1],\n [-1, 4.5]\n ]\n });\n net.states = tf.tensor1d([2, 3]);\n net.biases = tf.tensor1d([-2.75, -1.75]);\n\n net.run_duration = 100000;\n net.step_size = 1;\n net.name = \"Simple 2-Node oscillator.\";\n Object.freeze(net);\n return net;\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2036,"value":"// a 3x3 version with disconnected node 3\nfunction osc3NodeExample() {\n let net = ctNet({\n size: 3,\n init_weights: [\n // [3.99999999, 1, 1],\n [4.5, 1, 0.1],\n [-1, 4.5, 0],\n [1.1, 0.3, 1]\n ]\n });\n net.states = tf.tensor1d([2, 3, 100]);\n net.biases = tf.tensor1d([-2.75, -1.75, 0]);\n\n net.taus = tf.tensor1d([1, 1.1, 5]); //slow\n\n //net.run_duration = 1000;\n net.step_size = 1;\n net.name = \"Simple 3-Node.\";\n net.freeze(net, net.weights);\n // Object.freeze(net);\n return net;\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3511,"value":"// a 3x3 version with disconnected node 3\nfunction osc4NodeExample() {\n let net = ctNet({\n size: 4,\n init_weights: [\n // [3.99999999, 1, 1],\n [4.5001, 1, 0.1, 0],\n [-1, 4.5, 0, 0.2],\n [1.1, 0.3, 1, 0.6],\n [1.1, -0.5, 1, -0.1]\n ]\n });\n net.states = tf.tensor1d([2, 3, 10, 5]);\n net.biases = tf.tensor1d([-2.75, -1.75, 0, 1]);\n net.taus = tf.tensor1d([1, 1.1, 5, 2]); //slow\n\n //net.run_duration = 10000;\n net.step_size = 1;\n net.name = \"Simple 4-Node\";\n Object.freeze(net);\n return net;\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":4126,"value":"rand1d(9).arraySync()","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":4101,"value":"function rand1d(n, limit = 1) {\n return tf.tensor1d(Array.from({ length: n }, () => Math.random() * limit));\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":4156,"value":"coreloop.net.weights.arraySync()","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2481,"value":"// doesn't oscillate but 11C from same figure does, and this is the parameter set used in 1st example above.\n\nfunction Beer1995Fig4a3a() {\n let net = ctNet({\n size: 2,\n init_weights: [\n [5.25, 1],\n [-1, 5.25]\n ]\n });\n net.states = tf.tensor1d([20, 53]); // what here?\n net.biases = tf.tensor1d([-3.4, -2.5]);\n\n //net.run_duration = 3000;\n net.step_size = 1;\n net.name =\n \"Figure 4a3a from Beer 1995. Doesn't oscillate, but similar to those that do.\";\n Object.freeze(net);\n return net;\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3937,"value":"function Beer1995Example7() {\n let net = ctNet({\n size: 2,\n init_weights: [\n [6.3, 1],\n [-1, 6.3]\n ]\n });\n net.states = tf.tensor1d([20, 53]); // what here?\n net.biases = tf.tensor1d([-3.75, -2.65]);\n\n //net.run_duration = 3000;\n net.step_size = 1;\n net.name = \"7th example from Beer 1995. Try different inputs?\";\n Object.freeze(net);\n return net;\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":602,"value":"# ctlib as a draft of a library/api\n\n*Continuous Time (Recurrent Neural etc.) nets*\n\nstatus: not a sensible api, not tested, volatile.","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":4234,"value":"tf.setBackend(\"nodelgl\")","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":4232,"value":"tf.getBackend()","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":209,"value":"$47","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2005,"value":"async function* runSimulation(net) {\n let loopcount = 0;\n let livestates = net.states;\n let liveoutputs = net.outputs;\n var calc;\n while (loopcount++ < net.run_duration) {\n const calc = await net.eulerStep(net, livestates, liveoutputs);\n livestates = calc.states;\n liveoutputs = calc.outputs;\n\n yield {\n net: net,\n states: calc.states,\n states_cpu: calc.states.arraySync(),\n outputs: calc.outputs,\n outputs_cpu: calc.outputs.arraySync(),\n yprime: calc.yprime,\n yprime_cpu: calc.yprime.arraySync(),\n count: loopcount,\n in_fp: net.in_fixed_point(calc.yprime)\n };\n\n // OUCH\n // calc.outputs.dispose();\n // calc.yprime.dispose(); // for now flush the GPU version\n }\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":1642,"value":"pptensor = function pptensor(t) {\n return t.arraySync().map((x) => ppfloat(x));\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":1624,"value":"ppfloat = function ppfloat(x) {\n return Number.parseFloat(Number.parseFloat(x).toPrecision(2));\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":1997,"value":"function col(data, n) {\n return data.map(function (value, index) {\n return value[n];\n });\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2052,"value":"function changey(coreloop) {\n return Math.abs(\n ppfloat(coreloop.yprime.arraySync().reduce((a, b) => a + b, 0))\n );\n} // mean rate of activation change","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3613,"value":"# Analysis and plotting utility code","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3611,"value":"function makePlotMarks(loop) {\n let net = loop.net;\n let plotlines = [];\n let netPalette;\n const shuffle = d3.shuffler(d3.randomLcg(165));\n netPalette = shuffle(\n d3.quantize(d3.interpolateHcl(\"#f4e153\", \"#362142\"), net.size)\n );\n for (let ix = 0; ix < net.size; ix++) {\n plotlines.push(\n Plot.lineY(col(results, ix).slice(-400), {\n stroke: netPalette[ix]\n })\n );\n }\n return plotlines;\n}","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":2285,"value":"$48","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":1618,"value":"# ctlib construction API","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":60,"value":"class Monitor {\n constructor() {\n this._items = new Map();\n this._list = html` `;\n //\n\n const view = html`

${this._list}`;\n view.value = this;\n return view;\n }\n\n clear(prefix = \"\") {\n for (let [key, item] of this._items.entries()) {\n if (key.substr(0, prefix.length) !== prefix) continue;\n if (item.node.parentNode) item.node.parentNode.removeChild(item.node);\n this._items.delete(key);\n }\n }\n\n _initItem(name) {\n const label = html`${label}${value}`;\n this._list.appendChild(node);\n this._items.set(name, { label, value, node });\n }\n\n log(key, value) {\n if (!this._items.has(key)) this._initItem(key);\n return (this._items.get(key).value.textContent = value);\n }\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":649,"value":"tenNodeIdenticalWeights = ctNet({\n size: 10,\n init_weights: () => 0.2\n}).weights.arraySync()","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":660,"value":"TenNodesRandomSmallWeights = ctNet({\n size: 10,\n init_weights: () => Math.random() / 2\n}).weights.arraySync()","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":654,"value":"OscillatorWithoutBiases = ctNet({\n size: 2,\n init_weights: [\n [4.5, 1],\n [-1, 4.5]\n ]\n}).outputs.arraySync()","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":667,"value":"TwoByTwoWithInitialStates = {\n let x = ctNet(2);\n x.states = tf.tensor1d([2, 3]);\n return x;\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":777,"value":"/* HundredNodesLargeRandomWeights = ctNet({\n size: 100,\n init_weights: () => Math.random() * 1000\n}).weights.arraySync() */","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":821,"value":"# Utilities and Libraries","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":1059,"value":"tf = require(\"@tensorflow/tfjs@3.9.0\")","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":4050,"value":"import { div, print } from \"@visnup/hello-danfo-js\" // https://blog.tensorflow.org/2020/08/introducing-danfo-js-pandas-like-library-in-javascript.html","pinned":true,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":3568,"value":"function swatches(colors) {\n return html`${colors.map(\n (c) =>\n `

`\n )}`;\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"}],"files":"$W49","stdlib":"1"}} 45:"$44:body"

${DOM.text(name)}`;\n const value = html`

`;\n const node = html`