{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "This application demonstrates how to build a simple neural network using the Graph mark. \n", "Interactions can be enabled by adding event handlers (click, hover etc) on the nodes of the network. \n", "See the [Mark Interactions notebook](../Interactions/Mark%20Interactions.ipynb) and the [Scatter Notebook](../Marks/Object%20Model/Scatter.ipynb) for details." ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [], "source": [ "from itertools import chain, product\n", "\n", "import numpy as np\n", "from bqplot import LinearScale, Graph, Figure" ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [], "source": [ "class NeuralNet(Figure):\n", " def __init__(self, **kwargs):\n", " self.height = kwargs.get('height', 600)\n", " self.width = kwargs.get('width', 960)\n", " self.directed_links = kwargs.get('directed_links', False)\n", " \n", " self.num_inputs = kwargs['num_inputs']\n", " self.num_hidden_layers = kwargs['num_hidden_layers']\n", " self.nodes_output_layer = kwargs['num_outputs']\n", " self.layer_colors = kwargs.get('layer_colors', \n", " ['Orange'] * (len(self.num_hidden_layers) + 2))\n", " \n", " self.build_net()\n", " super(NeuralNet, self).__init__(**kwargs)\n", " \n", " def build_net(self):\n", " # create nodes\n", " self.layer_nodes = []\n", " self.layer_nodes.append(['x' + str(i+1) for i in range(self.num_inputs)])\n", " \n", " for i, h in enumerate(self.num_hidden_layers):\n", " self.layer_nodes.append(['h' + str(i+1) + ',' + str(j+1) for j in range(h)])\n", " self.layer_nodes.append(['y' + str(i+1) for i in range(self.nodes_output_layer)])\n", " \n", " self.flattened_layer_nodes = list(chain(*self.layer_nodes))\n", " \n", " # build link matrix\n", " i = 0\n", " node_indices = {}\n", " for layer in self.layer_nodes:\n", " for node in layer:\n", " node_indices[node] = i\n", " i += 1\n", "\n", " n = len(self.flattened_layer_nodes)\n", " self.link_matrix = np.empty((n,n))\n", " self.link_matrix[:] = np.nan\n", "\n", " for i in range(len(self.layer_nodes) - 1):\n", " curr_layer_nodes_indices = [node_indices[d] for d in self.layer_nodes[i]]\n", " next_layer_nodes = [node_indices[d] for d in self.layer_nodes[i+1]]\n", " for s, t in product(curr_layer_nodes_indices, next_layer_nodes):\n", " self.link_matrix[s, t] = 1\n", " \n", " # set node x locations\n", " self.nodes_x = np.repeat(np.linspace(0, 100, \n", " len(self.layer_nodes) + 1, \n", " endpoint=False)[1:], \n", " [len(n) for n in self.layer_nodes])\n", "\n", " # set node y locations\n", " self.nodes_y = np.array([])\n", " for layer in self.layer_nodes:\n", " n = len(layer)\n", " ys = np.linspace(0, 100, n+1, endpoint=False)[1:]\n", " self.nodes_y = np.append(self.nodes_y, ys[::-1])\n", " \n", " # set node colors\n", " n_layers = len(self.layer_nodes)\n", " self.node_colors = np.repeat(np.array(self.layer_colors[:n_layers]), \n", " [len(layer) for layer in self.layer_nodes]).tolist()\n", " \n", " xs = LinearScale(min=0, max=100)\n", " ys = LinearScale(min=0, max=100)\n", " \n", " self.graph = Graph(node_data=[{'label': d, \n", " 'label_display': 'none'} for d in self.flattened_layer_nodes], \n", " link_matrix=self.link_matrix, \n", " link_type='line',\n", " colors=self.node_colors,\n", " directed=self.directed_links,\n", " scales={'x': xs, 'y': ys}, \n", " x=self.nodes_x, \n", " y=self.nodes_y,\n", " # color=2 * np.random.rand(len(self.flattened_layer_nodes)) - 1\n", " )\n", " self.graph.hovered_style = {'stroke': '1.5'}\n", " self.graph.unhovered_style = {'opacity': '0.4'}\n", " \n", " self.graph.selected_style = {'opacity': '1',\n", " 'stroke': 'red',\n", " 'stroke-width': '2.5'}\n", " self.marks = [self.graph]\n", " self.title = 'Neural Network'\n", " self.layout.width = str(self.width) + 'px'\n", " self.layout.height = str(self.height) + 'px'" ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [ { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "b2d75468c39d4bcf896561f72febb74d", "version_major": 2, "version_minor": 0 }, "text/plain": [ "NeuralNet(fig_margin={'top': 60, 'bottom': 60, 'left': 60, 'right': 60}, layout=Layout(height='600px', width='…" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "NeuralNet(num_inputs=3, num_hidden_layers=[10, 10, 8, 5], num_outputs=1)" ] } ], "metadata": { "input_collapsed": false, "kernelspec": { "display_name": "Python 3", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.7.6" } }, "nbformat": 4, "nbformat_minor": 2 }