Updates figures from chapter 4 and 5

plots/balance_stdp.R

@@ -95,7 +95,8 @@ plot_weights <- function(df_data, bins, bin_width, subset=NULL){
     w_distribution <- df_weights %>%
         ggplot(aes(x=mids, y=count)) +
         geom_col(width=bin_width, fill=color_map[2], color=color_map[2]) + 
-        theme_bw() + labs(x='weights (pA)', y='fraction of synapses')
+        theme_bw() + labs(x='weights (pA)', y='fraction of synapses') +
+        theme(text=element_text(size=16))
 
     return(w_distribution)
 }
@@ -110,8 +111,10 @@ protocols <- map_int(metadata, \(x) x$protocol)
 prot_id <- match(1, protocols)
 event_data <- read.csv(file.path(data_path[prot_id], 'events_spikes.csv'))
 fetches <- event_data %>%
+    mutate(time_ms = time_ms/1000) %>%
     ggplot(aes(x=time_ms, y=num_events)) + geom_line() +
-    theme_bw() + labs(x='time (ms)', y='# Active neurons')
+    theme_bw() + labs(x='time (s)', y='# active neurons') +
+    theme(text=element_text(size=16))
 
 w_distribution <- plot_weights(data_path[prot_id], bins=19, bin_width=0.64)
 incoming_w_j0 <- plot_weights(data_path[prot_id], bins=19, bin_width=0.64, 0)
@@ -137,8 +140,10 @@ if(metadata[[prot_id[1]]]$alpha==.1449) {
 event_data <- read.csv(file.path(data_path[prot_id], 'events_spikes.csv'))
 fetches <- event_data %>%
     filter(time_ms < 195000) %>%
+    mutate(time_ms = time_ms/1000) %>%
     ggplot(aes(x=time_ms, y=num_events)) + geom_line() +
-    theme_bw() + labs(x='time (ms)', y='# Active neurons')
+    theme_bw() + labs(x='time (s)', y='# active neurons') +
+    theme(text=element_text(size=16))
 
 w_distribution <- plot_weights(data_path[prot_id], bins=35, bin_width=.7)
 w_distribution <- w_distribution +
@@ -160,7 +165,8 @@ df_rates_init$group <- 'initial'
 freq <- bind_rows(df_rates_init, df_rates_final) %>%
         ggplot(aes(x=rate, color=group, fill=group)) + theme_bw() +
         geom_histogram(alpha=.5) + scale_color_manual(values=color_map) + 
-        scale_fill_manual(values=color_map) + theme(legend.position = c(0.8, 0.8)) +
+        scale_fill_manual(values=color_map) +
+        theme(legend.position = c(0.8, 0.8), text=element_text(size=16)) +
         labs(x='firing rates (Hz)', y='count', fill=element_blank(), color=element_blank()) +
         scale_y_log10()
 

plots/conditions_stdp.R

@@ -52,42 +52,45 @@ t_final <- 250
 fetch_dist_init <- df_events %>%
     filter(time_ms > t_init & time_ms < t_final) %>%
     ggplot(aes(x=group, y=num_fetch, fill=group)) +
-    geom_boxplot(outlier.shape=NA) + theme_bw() + theme(legend.position="none") +
+    geom_boxplot(outlier.shape=NA) + theme_bw() +
+    theme(legend.position="none", text = element_text(size=16)) +
     geom_jitter(color="black", size=0.7, alpha=0.5) +
-    labs(x=element_blank(), y='# Memory access') +
+    labs(x=element_blank(), y='mems') +
     scale_fill_manual(values=color_map)
 t_init <- 0
 t_final <- 100
 trace_init <- df_events %>%
     filter(time_ms > t_init & time_ms < t_final) %>%
     ggplot(aes(x=time_ms, y=num_fetch, color=group)) + geom_line() +
-    theme(legend.position="none") + theme_bw() +
+    theme_bw() + theme(legend.position="none", text = element_text(size=16)) +
     guides(color=guide_legend(override.aes=list(linewidth=4))) +
-    labs(x='time (ms)', y='# Memory access', color='Event threshold') +
+    labs(x='time (ms)', y='mems', color='Event threshold') +
     scale_color_manual(values=color_map)
 
 t_init <- 99500
 t_final <- 99750
 fetch_dist_final <- df_events %>%
     filter(time_ms > t_init & time_ms < t_final) %>%
     ggplot(aes(x=group, y=num_fetch, fill=group)) +
-    geom_boxplot(outlier.shape=NA) + theme_bw() + theme(legend.position="none") +
+    geom_boxplot(outlier.shape=NA) + theme_bw() +
+    theme(legend.position="none", text = element_text(size=16)) +
     geom_jitter(color="black", size=0.7, alpha=0.5) +
-    labs(x=element_blank(), y='# Memory access') +
+    labs(x=element_blank(), y='mems') +
     scale_fill_manual(values=color_map)
 t_init <- 99500
 t_final <- 100000
 trace_final <- df_events %>%
     filter(time_ms > t_init & time_ms < t_final) %>%
     ggplot(aes(x=time_ms, y=num_fetch, color=group)) + geom_line() +
-    theme_bw() + labs(x='time (ms)', y='# Memory access') +
+    theme_bw() + labs(x='time (ms)', y='mems') +
     scale_x_continuous(labels=c(t_init, t_final), breaks=c(t_init, t_final)) +
-    scale_color_manual(values=color_map) + theme(legend.position="none")
+    scale_color_manual(values=color_map) +
+    theme(legend.position="none", text = element_text(size=16))
 
 df_spk <- read.csv(str_replace(nth(dir_list, -1), "events_spikes.feather", "spikes_post.csv"))
 rates <- df_spk %>%
     ggplot(aes(x=time_ms)) + geom_histogram(fill=color_map[3]) + theme_bw() +
-    labs(x='time (ms)', y='# spikes')
+    labs(x='time (ms)', y='# spikes') + theme(text = element_text(size=16))
 
 n_fetch <- rates /
        (trace_init + trace_final) /

plots/fp8_simple_stdp.R

@@ -41,10 +41,10 @@ state_vars <- df_vars %>%
     filter(!str_detect(monitor, "ref"), id==0) %>%
     mutate(monitor = str_replace(monitor,
                                  "statemon_pre_neurons",
-                                 "presynaptic neuron")) %>%
+                                 "presynaptic")) %>%
     mutate(monitor = str_replace(monitor,
                                  "statemon_post_neurons",
-                                 "postsynaptic neuron")) %>%
+                                 "postsynaptic")) %>%
     mutate(variable = str_replace(variable, "Ca", "x")) %>%
     mutate(variable = str_replace(variable, "g", "PSP"))
 state_vars$value <- map_dbl(state_vars$value, minifloat2decimal)
@@ -54,7 +54,7 @@ states_stoch <- state_vars %>%
     geom_line() + labs(x='time (ms)', y='magnitude (a.u.)', color='state variables') +
     xlim(c(363, 713)) + scale_color_manual(values=color_map[c(1, 3, 5)]) +
     guides(color=guide_legend(override.aes=list(linewidth=5))) +
-    theme(legend.position="none")
+    theme(legend.position="none", text = element_text(size=16))
 
 df_weights <- read.csv(file.path(data_path[sim_id], 'synapse_vars.csv'))
 w_trace <- df_weights %>%
@@ -71,7 +71,7 @@ weights_stoch_high <- temp_weights +
     geom_line(data=avg_weight, aes(x=time_ms, y=value),
               color=color_map[4], linewidth=1) +
     labs(x='time (ms)', y='weight (a.u.)') +
-    xlim(c(363, 1957)) + theme_bw()
+    xlim(c(363, 1957)) + theme_bw() + theme(text=element_text(size=16))
 
 selected_folder <- stoch_round & small_w
 sim_id <- match(T, selected_folder)
@@ -90,7 +90,7 @@ weights_stoch_low <- temp_weights +
     geom_line(data=avg_weight, aes(x=time_ms, y=value),
               color=color_map[4], linewidth=1) +
     labs(x='time (ms)', y='weight (a.u.)') +
-    xlim(c(363, 1957)) + theme_bw()
+    xlim(c(363, 1957)) + theme_bw() + theme(text=element_text(size=16))
 
 selected_folder <- !stoch_round & !small_w
 sim_id <- match(T, selected_folder)
@@ -100,10 +100,10 @@ state_vars <- df_vars %>%
     filter(!str_detect(monitor, "ref"), id==0) %>%
     mutate(monitor = str_replace(monitor,
                                  "statemon_pre_neurons",
-                                 "presynaptic neuron")) %>%
+                                 "presynaptic")) %>%
     mutate(monitor = str_replace(monitor,
                                  "statemon_post_neurons",
-                                 "postsynaptic neuron")) %>%
+                                 "postsynaptic")) %>%
     mutate(variable = str_replace(variable, "Ca", "x")) %>%
     mutate(variable = str_replace(variable, "g", "PSP"))
 state_vars$value <- map_dbl(state_vars$value, minifloat2decimal)
@@ -112,7 +112,8 @@ states_det <- state_vars %>%
     facet_grid(rows=vars(fct_rev(monitor)), scales="free") +
     geom_line() + labs(x='time (ms)', y='magnitude (a.u.)', color='state variables') +
     xlim(c(363, 713)) + scale_color_manual(values=color_map[c(1, 3, 5)]) +
-    guides(color=guide_legend(override.aes=list(linewidth=5)))
+    guides(color=guide_legend(override.aes=list(linewidth=5))) +
+    theme(text=element_text(size=16))
 
 df_weights <- read.csv(file.path(data_path[sim_id], 'synapse_vars.csv'))
 w_trace <- df_weights %>%
@@ -122,6 +123,7 @@ weights_det_high <- w_trace %>%
     filter(id==0) %>%
     ggplot(aes(x=time_ms, y=value, group=id)) +
     geom_line(alpha=.6, color=color_map[3]) + theme_bw() +
+    theme(text=element_text(size=16)) +
     scale_color_manual(values = color_map[3]) +
     labs(x='time (ms)', y='weight (a.u.)') +
     xlim(c(363, 1957))
@@ -137,6 +139,7 @@ weights_det_low <- w_trace %>%
     filter(id==0) %>%
     ggplot(aes(x=time_ms, y=value, group=id)) +
     geom_line(alpha=.6, color=color_map[3]) + theme_bw() +
+    theme(text=element_text(size=16)) +
     scale_color_manual(values = color_map[3]) +
     labs(x='time (ms)', y='weight (a.u.)') +
     xlim(c(363, 1957))

plots/multi_fetch_stdp.R

@@ -57,7 +57,8 @@ df_events <- df_events %>%
     group_by(N_post, group) %>% mutate(mean_fetch=mean(num_fetch))
 multi_fetch_avg <- df_events %>%
     ggplot(aes(x=group, y=num_fetch)) +
-    geom_boxplot(aes(fill=group)) + theme_bw() + theme(legend.position="none") +
+    geom_boxplot(aes(fill=group)) + theme_bw() +
+    theme(legend.position="none", text=element_text(size=16)) +
     geom_point(aes(y=mean_fetch, color=group), shape=2, size=7) +
     facet_wrap(~N_post) + scale_y_log10() +
     labs(x=element_blank(), y='# Memory access') +

plots/simple_stdp.R

@@ -27,15 +27,16 @@ ca_traces <- ggplot() +
     geom_histogram(data=df_pre_events, aes(x=time_ms), fill=color_map[2],
                    alpha=0.3, binwidth=1) +
     labs(x=element_blank(), y=TeX(r'($x_i$ (a.u))')) +
-    theme_bw() + theme(panel.grid.minor=element_blank())
+    theme_bw() + theme(panel.grid.minor=element_blank(), text = element_text(size=16))
 
 df_post_ca_trace <- df_vars %>%
     filter(id==0) %>%
     filter(variable=='Ca' & monitor=='statemon_post_neurons') %>%
     filter(time_ms>690 & time_ms<890) %>%
     ggplot(aes(x=time_ms, y=value)) +
     labs(x='time (ms)', y=TeX(r'($x_j$ (a.u))')) +
-    theme_bw() + theme(panel.grid.minor=element_blank()) + geom_line()
+    theme_bw() + theme(panel.grid.minor=element_blank(), text = element_text(size=16)) +
+    geom_line()
 
 b_rang <- c(0, 310, 420, 720, 800, 1110, 1210, 1510, 1610, 2010, 2080, 2320)
 w_trace <- df_weights %>%
@@ -57,7 +58,8 @@ w_trace <- df_weights %>%
     annotate(geom='text', x=b_rang[6] - (b_rang[6]-b_rang[5])/2, y=11.2, label='III') +
     annotate(geom='text', x=b_rang[8] - (b_rang[8]-b_rang[7])/2, y=11.2, label='IV') +
     annotate(geom='text', x=b_rang[10] - (b_rang[10]-b_rang[9])/2, y=11.2, label='V') +
-    annotate(geom='text', x=b_rang[12] - (b_rang[12]-b_rang[11])/2, y=11.2, label='VI')
+    annotate(geom='text', x=b_rang[12] - (b_rang[12]-b_rang[11])/2, y=11.2, label='VI') +
+    theme(text = element_text(size=16))
 
 # Mean is taken over all connections, i.e. id
 mse_w <- df_weights %>%
@@ -68,10 +70,12 @@ mse_w <- df_weights %>%
     group_by(time_ms) %>%
     summarise(mse = mean(squared_error)) %>%
     ggplot(aes(x=time_ms, y=mse)) + labs(x='time (ms)', y='mean squared error') +
-    geom_line() + theme_bw() + theme(panel.grid.minor=element_blank())
+    geom_line() + theme_bw() +
+    theme(panel.grid.minor=element_blank(), text = element_text(size=16)) +
+    scale_x_continuous(breaks = scales::pretty_breaks(n = 3))
 
-w_changes <- wrap_elements(w_trace + plot_annotation(title='A')) / 
-    (wrap_elements(mse_w + plot_annotation(title='B')) +
-     wrap_elements(ca_traces / df_post_ca_trace + plot_annotation(title='C')))
+w_changes <- wrap_elements(w_trace + plot_annotation(title='A', theme=theme(text=element_text(size=16)))) / 
+    (wrap_elements(mse_w + plot_annotation(title='B', theme=theme(text=element_text(size=16)))) +
+     wrap_elements(ca_traces / df_post_ca_trace + plot_annotation(title='C', theme=theme(text=element_text(size=16)))))
 
 ggsave(argv$dest, w_changes)

plots/stochastic_balance_stdp.R

@@ -96,7 +96,8 @@ plot_weights <- function(df_data, bin_width, subset=NULL){
     w_distribution <- df_weights %>%
         ggplot(aes(x=mids, y=count)) +
         geom_col(width=bin_width, fill=color_map[1], color=color_map[1]) + 
-        theme_bw() + labs(x='weights (a.u.)', y='fraction of synapses')
+        theme_bw() + labs(x='weights (a.u.)', y='fraction of synapses') +
+        theme(text=element_text(size=16))
 
     return(w_distribution)
 }
@@ -116,18 +117,20 @@ variables <- state_vars %>%
     mutate(variable = str_replace(variable, "Ca", "x")) %>%
     mutate(variable = str_replace(variable, "g", "PSP")) %>%
     ggplot(aes(x=time_ms, y=value, color=variable)) +
-    labs(x='time (s)', y='magnitude (a.u.)') + scale_color_manual(values=color_map) +
+    labs(x='time (ms)', y='magnitude (a.u.)') + scale_color_manual(values=color_map) +
     guides(color=guide_legend(override.aes=list(linewidth=5))) +
-    geom_line() + theme_bw()
+    geom_line() + theme_bw() + theme(text=element_text(size=16))
 
 w_distribution <- plot_weights(data_path, bin_width=0.64)
 incoming_w_j0 <- plot_weights(data_path, bin_width=0.64, 0)
 incoming_w_j10 <- plot_weights(data_path, bin_width=0.64, 10000)
 
 event_data <- read.csv(file.path(data_path, 'events_spikes.csv'))
 fetches <- event_data %>%
+    mutate(time_ms = time_ms/1000) %>%
     ggplot(aes(x=time_ms, y=num_events)) + geom_line(color=color_map[1]) +
-    theme_bw() + labs(x='time (ms)', y='# active neurons')
+    theme_bw() + labs(x='time (s)', y='# active neurons') +
+    theme(text=element_text(size=16))
 
 df_raster <- read.csv(file.path(data_path, 'output_spikes.csv'))
 print("Statistics with bimodal experiment: ")
@@ -146,10 +149,13 @@ df_rates_final$group <- 'final'
 df_rates_init$group <- 'initial'
 freq <- bind_rows(df_rates_init, df_rates_final) %>%
         ggplot(aes(x=rate, color=group, fill=group)) + theme_bw() +
-        geom_histogram(alpha=.5) + scale_color_manual(values=color_map) + 
-        scale_fill_manual(values=color_map) + theme(legend.position = c(0.8, 0.8)) +
-        labs(x='firing rates (Hz)', y='count', fill=element_blank(), color=element_blank()) +
-        scale_y_log10()
+        geom_histogram(alpha=.5, binwidth=1) + scale_color_manual(values=color_map) + 
+        scale_fill_manual(values=color_map) +
+        scale_x_continuous(breaks = scales::pretty_breaks(n = 4)) +
+        theme(legend.position = c(0.25, 0.8),
+              text=element_text(size=16),
+              legend.text=element_text(size=10)) +
+        labs(x='firing rates (Hz)', y='count', fill=element_blank(), color=element_blank())
 
 fig <- fetches + variables + freq + w_distribution +
     plot_annotation(tag_levels='A')

plots/stochastic_kernel.R

@@ -37,7 +37,7 @@ generate_plot <- function(df, time) {
       ggplot(aes(x=id, y=value)) + geom_line(color=color_map[3]) + theme_bw() +
       labs(x=TeX(r'($\Delta$\;t (ms))'), y=element_blank(), color=element_blank()) +
       guides(color=guide_legend(override.aes=list(linewidth=4))) +
-      theme(legend.position = c(0.25, 0.8))
+      theme(legend.position = c(0.25, 0.8), text=element_text(size=16))
   return(kernel)
 }
 
@@ -61,8 +61,8 @@ kernel_small_middle <- kernel_small_middle + labs(y='weight (a.u.)')
 kernel_small_final <- generate_plot(df_weights_small, 9000)
 
 fig_kernel <- wrap_elements(kernel_high_middle + kernel_high_final +
-                            plot_annotation(title='A')) /
+                            plot_annotation(title='A', theme=theme(text=element_text(size=16)))) /
               wrap_elements(kernel_small_middle + kernel_small_final +
-                            plot_annotation(title='B'))
+                            plot_annotation(title='B', theme=theme(text=element_text(size=16))))
 
 ggsave(argv$dest, fig_kernel)

plots/stochastic_operations.R

@@ -63,7 +63,8 @@ if (op_type == "addition"){
     L_heatmap <- ggplot(df_L, aes(delta, lambda, fill=L)) +
         scale_fill_gradientn(colors=color_map, name=TeX(r'($L$)')) +
         labs(x=TeX(r'($\delta$)'), y=TeX(r'($\lambda$)')) + theme_bw() + geom_tile() +
-        theme(legend.position="top")
+        theme(legend.position="top", text = element_text(size=16),
+              legend.text=element_text(angle=45))
 }
 
 my_breaks<-c(1, 10, 100, 1000)
@@ -76,7 +77,8 @@ probs_scatter <- ggplot(df_scatter, aes(x=p_round_closer_ideal, y=p_round_closer
     geom_smooth(colour='black', method=lm) + geom_bin2d(bins=25) +
     geom_abline(slope=1, intercept=0, linetype="dashed", color="black") +
     theme_bw() + labs(x=TeX(r'($P_i$)'), y=TeX(r'($P_s$)')) +
-    theme(legend.position="top")
+    theme(legend.position="top", text = element_text(size=16),
+          legend.text=element_text(angle=45))
 
 if (op_type == "addition"){
     sim_id <- match("stickier", selected_folders)
@@ -90,7 +92,7 @@ if (op_type == "addition"){
         geom_smooth(colour='black', method=lm) + geom_bin2d(bins=25) +
         geom_abline(slope=1, intercept=0, linetype="dashed", color="black") +
         theme_bw() + labs(x=TeX(r'($P_i$)'), y=TeX(r'($P_s$)')) +
-        theme(legend.position="none")
+        theme(legend.position="none", text = element_text(size=16))
 
     sim_id <- match("cutoff", selected_folders)
     df_cut <- read.csv(file.path(data_path[sim_id], "probs.csv"))
@@ -103,7 +105,7 @@ if (op_type == "addition"){
         geom_smooth(colour='black', method=lm) + geom_bin2d(bins=25) +
         geom_abline(slope=1, intercept=0, linetype="dashed", color="black") +
         theme_bw() + labs(x=TeX(r'($P_i$)'), y=TeX(r'($P_s$)')) +
-        theme(legend.position="none")
+        theme(legend.position="none", text = element_text(size=16))
 }
 
 if (op_type == "addition"){

plots/stochastic_pd.R

@@ -42,7 +42,8 @@ freq <- rates %>%
     mutate(layer=fct_rev(layer)) %>%
     ggplot(aes(rate, layer)) + theme_bw() +
     theme(panel.grid.minor=element_blank(),
-          panel.grid.major=element_blank()) +
+          panel.grid.major=element_blank(),
+          text=element_text(size=16)) +
     labs(x='firing rates (Hz)', y=element_blank()) +
     geom_boxplot() +
     stat_summary(fun='mean', geom='point', shape=2)
@@ -123,7 +124,7 @@ histograms <- df_raster %>%
     ggplot(aes(x=t, color=type)) + geom_step(stat='bin', bins=50) +
     guides(color=guide_legend(override.aes=list(size=4))) +
     facet_grid(rows=vars(laminae), scales='free') + theme_bw() +
-    theme(legend.position = 'bottom') +
+    theme(legend.position = 'bottom', text=element_text(size=16)) +
     scale_color_manual(values=color_map[c(1, 3)]) +
     guides(color=guide_legend(override.aes=list(linewidth=5))) +
     labs(x='time (ms)') + xlim(670, 730)
@@ -136,7 +137,9 @@ variables <- state_vars %>%
     ggplot(aes(x=time_ms, y=value, color=variable)) +
     geom_line() + theme_bw() +
     facet_grid(layer ~ type, scales="free") +
-    theme(legend.position = 'bottom', strip.text.y = element_blank()) +
+    theme(legend.position = 'bottom',
+          strip.text.y = element_blank(),
+          text=element_text(size=16)) +
     labs(x='time (s)', y='magnitude (a.u.)') + scale_color_manual(values=color_map) +
     guides(color=guide_legend(override.aes=list(linewidth=5)))