In EagerAI/fastai: Interface to 'fastai'

knitr::opts_chunk$set(echo = TRUE,eval = FALSE,echo = T)

Intro

First, we need to install timeseries_fastai module for Time-Series integration.

reticulate::py_install('git+https://github.com/tcapelle/timeseries_fastai.git', pip = TRUE)

Grab data

Download the data from Facebook Github page (Prophet):

library(dplyr)
library(fastai)

df = data.table::fread('https://raw.githubusercontent.com/facebook/prophet/master/examples/example_wp_log_peyton_manning.csv')

           ds        y
1: 2007-12-10 9.590761
2: 2007-12-11 8.519590
3: 2007-12-12 8.183677
4: 2007-12-13 8.072467
5: 2007-12-14 7.893572
6: 2007-12-15 7.783641

Preprocess

Scale and fill NA's:

split_idx = which(df$ds=='2016-01-01') # take 1 year for validation

y = df$y

df = timetk::tk_augment_timeseries_signature(df) %>%
  mutate_if(is.factor,  as.numeric) %>%
  select(-ds, -hour, -minute, -second, -hour12, -am.pm, -y) %>%
  scale() %>% data.table::as.data.table()

df[is.na(df)]=0
df$y = y

Split

Split data into 2 parts:

• train
• test

df_train = df[1:split_idx,]
df_test = df[(split_idx+1):nrow(df),]

x_cols = setdiff(colnames(df_train),'y')

Dataloader and Inception model

dls = TSDataLoaders_from_dfs(df_train, df_test, x_cols = x_cols, label_col = 'y', bs=60,
                             y_block = RegressionBlock())

dls %>% show_batch()

inception = create_inception(1, 1)

learn = Learner(dls, inception, metrics=list(mae(), rmse()))

Sequential(
  (0): SequentialEx(
    (layers): ModuleList(
      (0): InceptionModule(
        (convs): ModuleList(
          (0): Conv1d(1, 32, kernel_size=(39,), stride=(1,), padding=(19,), bias=False)
          (1): Conv1d(1, 32, kernel_size=(19,), stride=(1,), padding=(9,), bias=False)
          (2): Conv1d(1, 32, kernel_size=(9,), stride=(1,), padding=(4,), bias=False)
        )
        (conv_bottle): Sequential(
          (0): MaxPool1d(kernel_size=3, stride=1, padding=1, dilation=1, ceil_mode=False)
          (1): Conv1d(1, 32, kernel_size=(1,), stride=(1,), bias=False)
        )
        (bn_relu): Sequential(
          (0): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU()
        )
      )
    )
  )
  (1): SequentialEx(
    (layers): ModuleList(
      (0): InceptionModule(
        (bottleneck): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        (convs): ModuleList(
          (0): Conv1d(32, 32, kernel_size=(39,), stride=(1,), padding=(19,), bias=False)
          (1): Conv1d(32, 32, kernel_size=(19,), stride=(1,), padding=(9,), bias=False)
          (2): Conv1d(32, 32, kernel_size=(9,), stride=(1,), padding=(4,), bias=False)
        )
        (conv_bottle): Sequential(
          (0): MaxPool1d(kernel_size=3, stride=1, padding=1, dilation=1, ceil_mode=False)
          (1): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        )
        (bn_relu): Sequential(
          (0): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU()
        )
      )
    )
  )
  (2): SequentialEx(
    (layers): ModuleList(
      (0): InceptionModule(
        (bottleneck): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        (convs): ModuleList(
          (0): Conv1d(32, 32, kernel_size=(39,), stride=(1,), padding=(19,), bias=False)
          (1): Conv1d(32, 32, kernel_size=(19,), stride=(1,), padding=(9,), bias=False)
          (2): Conv1d(32, 32, kernel_size=(9,), stride=(1,), padding=(4,), bias=False)
        )
        (conv_bottle): Sequential(
          (0): MaxPool1d(kernel_size=3, stride=1, padding=1, dilation=1, ceil_mode=False)
          (1): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        )
        (bn_relu): Sequential(
          (0): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU()
        )
      )
      (1): Shortcut(
        (act_fn): ReLU(inplace=True)
        (conv): ConvLayer(
          (0): Conv1d(128, 128, kernel_size=(1,), stride=(1,), bias=False)
          (1): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU()
        )
        (bn): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
  )
  (3): SequentialEx(
    (layers): ModuleList(
      (0): InceptionModule(
        (bottleneck): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        (convs): ModuleList(
          (0): Conv1d(32, 32, kernel_size=(39,), stride=(1,), padding=(19,), bias=False)
          (1): Conv1d(32, 32, kernel_size=(19,), stride=(1,), padding=(9,), bias=False)
          (2): Conv1d(32, 32, kernel_size=(9,), stride=(1,), padding=(4,), bias=False)
        )
        (conv_bottle): Sequential(
          (0): MaxPool1d(kernel_size=3, stride=1, padding=1, dilation=1, ceil_mode=False)
          (1): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        )
        (bn_relu): Sequential(
          (0): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU()
        )
      )
    )
  )
  (4): SequentialEx(
    (layers): ModuleList(
      (0): InceptionModule(
        (bottleneck): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        (convs): ModuleList(
          (0): Conv1d(32, 32, kernel_size=(39,), stride=(1,), padding=(19,), bias=False)
          (1): Conv1d(32, 32, kernel_size=(19,), stride=(1,), padding=(9,), bias=False)
          (2): Conv1d(32, 32, kernel_size=(9,), stride=(1,), padding=(4,), bias=False)
        )
        (conv_bottle): Sequential(
          (0): MaxPool1d(kernel_size=3, stride=1, padding=1, dilation=1, ceil_mode=False)
          (1): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        )
        (bn_relu): Sequential(
          (0): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU()
        )
      )
    )
  )
  (5): SequentialEx(
    (layers): ModuleList(
      (0): InceptionModule(
        (bottleneck): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        (convs): ModuleList(
          (0): Conv1d(32, 32, kernel_size=(39,), stride=(1,), padding=(19,), bias=False)
          (1): Conv1d(32, 32, kernel_size=(19,), stride=(1,), padding=(9,), bias=False)
          (2): Conv1d(32, 32, kernel_size=(9,), stride=(1,), padding=(4,), bias=False)
        )
        (conv_bottle): Sequential(
          (0): MaxPool1d(kernel_size=3, stride=1, padding=1, dilation=1, ceil_mode=False)
          (1): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        )
        (bn_relu): Sequential(
          (0): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU()
        )
      )
      (1): Shortcut(
        (act_fn): ReLU(inplace=True)
        (conv): ConvLayer(
          (0): Conv1d(128, 128, kernel_size=(1,), stride=(1,), bias=False)
          (1): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (2): ReLU()
        )
        (bn): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
  )
  (6): AdaptiveConcatPool1d(
    (ap): AdaptiveAvgPool1d(output_size=1)
    (mp): AdaptiveMaxPool1d(output_size=1)
  )
  (7): Flatten(full=False)
  (8): Linear(in_features=256, out_features=1, bias=True)
)

Optimal LR

Find suitable learning rate and plot:

lrs = learn %>% lr_find()

learn %>% plot_lr_find()

Conclusion

Fit and include early stopping:

learn %>% fit_one_cycle(30, 1e-5, cbs = EarlyStoppingCallback(patience = 5))

learn %>% predict(df_test)

          y
1  5.925275
2  5.925579
3  6.060019
4  5.749205
5  5.793755
6  5.974738
7  7.185201
8  6.897375
9  7.095043
10 7.082201
11 6.916879
12 6.802319
13 6.714959
14 6.589868
15 6.436776
16 7.105388
17 7.237350
18 7.133458
19 7.103930

EagerAI/fastai documentation built on April 16, 2024, 12:01 p.m.